Conflict Arguments Between R and Python: Error in py_call_impl(callable, call_args$unnamed, call_args$named)
Waselia opened this issue · 20 comments
Waselia commented
Hi all,
I am trying to run a code published by: Gabriella Andrea. I use the provided dataset that comes along with the code. I start with the NeuralNetworkModels file, works just fine. Then I continue with the EmbeddingWeights file, and the problem begins here.
Details about the error will be shown below. I did try to reinstall R and Python, and updated all relevant packages. From what I understand about the error, there might seems to be a conflict between the R code for the layer_embedding command, and the embedding command from Python. I didn't change any of the original code, and from what I know, the command follows the correct arguments and usage as described here.
Please help, I have been stuck on this error for almost a week now.
Starting from the "Application of Neural Network Model" part (line 177) of the EmbeddingWeights file, I received the following error:
> model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)
2024-04-06 23:17:44.652965: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-04-06 23:17:56.164768: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2024-04-06 23:18:21.410172: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
ValueError: Unrecognized keyword arguments passed to Embedding: {'input_length': 1, 'weights': [array([[1.]])]}
Run `reticulate::py_last_error()` for details.I followed the instruction and tried the reticulate::py_last_error() code:
> reticulate::py_last_error()
── Python Exception Message ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Traceback (most recent call last):
File "C:\Users\Admin\OneDrive\DOCUME~1\VIRTUA~1\R-RETI~1\Lib\site-packages\keras\src\layers\core\embedding.py", line 81, in __init__
super().__init__(**kwargs)
File "C:\Users\Admin\OneDrive\DOCUME~1\VIRTUA~1\R-RETI~1\Lib\site-packages\keras\src\layers\layer.py", line 264, in __init__
raise ValueError(
ValueError: Unrecognized keyword arguments passed to Embedding: {'input_length': 1, 'weights': [array([[1.]])]}
── R Traceback ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
▆
1. ├─global neural_network_train_embedding(...)
2. │ └─... %>% layer_flatten
3. ├─keras::layer_flatten(.)
4. │ └─keras::create_layer(keras$layers$Flatten, object, args)
5. ├─keras::layer_embedding(...)
6. │ └─keras::create_layer(keras$layers$Embedding, object, args)
7. │ ├─base::do.call(layer_class, args)
8. │ └─reticulate (local) `<python.builtin.type>`(...)
9. │ └─reticulate:::py_call_impl(callable, call_args$unnamed, call_args$named)
10. ├─base::tryCatch(...)
11. │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
12. │ ├─base (local) tryCatchOne(...)
13. │ │ └─base (local) doTryCatch(return(expr), name, parentenv, handler)
14. │ └─base (local) tryCatchList(expr, names[-nh], parentenv, handlers[-nh])
15. │ └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
16. │ └─base (local) doTryCatch(return(expr), name, parentenv, handler)
17. └─base::evalq(`<fn>`(TRUE, 1L), `<env>`)
18. └─base::evalq(`<fn>`(TRUE, 1L), `<env>`)
See `reticulate::py_last_error()$r_trace$full_call` for more details.Here's a more detailed explanation I receive from the error:
> reticulate::py_last_error()$r_trace$full_call
[[1]]
neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)
[[2]]
dropout_embed <- dropout %>%
layer_embedding(input_dim = 1, output_dim = 1, trainable=FALSE, input_length = 1,
weights=list(matrix(c(1)))) %>%
layer_flatten
[[3]]
layer_flatten(.)
[[4]]
create_layer(keras$layers$Flatten, object, args)
[[5]]
layer_embedding(., input_dim = 1, output_dim = 1, trainable = FALSE,
input_length = 1, weights = list(matrix(c(1))))
[[6]]
create_layer(keras$layers$Embedding, object, args)
[[7]]
do.call(layer_class, args)
[[8]]
(function (input_dim, output_dim, embeddings_initializer = "uniform",
embeddings_regularizer = NULL, embeddings_constraint = NULL,
mask_zero = FALSE, lora_rank = NULL, ...)
{
cl <- sys.call()
cl[[1L]] <- list2
call_args <- split_named_unnamed(eval(cl, parent.frame()))
result <- py_call_impl(callable, call_args$unnamed, call_args$named)
if (py_get_convert(callable))
result <- py_to_r(result)
if (is.null(result))
invisible(result)
else result
})(input_dim = 1L, output_dim = 1L, input_length = 1L, trainable = FALSE,
weights = list(1))
[[9]]
py_call_impl(callable, call_args$unnamed, call_args$named)
[[10]]
tryCatch(evalq(function (maybe_use_cached = FALSE, trim_tail = 1L)
{
t <- rlang::trace_back()
t <- t[1L:(nrow(t) - trim_tail), ]
t$full_call <- sys.calls()[seq_len(nrow(t))]
if (!maybe_use_cached)
return((.globals$last_r_trace <- t))
ot <- .globals$last_r_trace
if (is.null(ot) || nrow(t) >= nrow(ot) || !identical(t, ot[seq_len(nrow(t)),
])) {
.globals$last_r_trace <- t
}
invisible(.globals$last_r_trace)
}(TRUE, 1L), <environment>), error = function (x)
x, interrupt = function (x)
x)
[[11]]
tryCatchList(expr, classes, parentenv, handlers)
[[12]]
tryCatchOne(tryCatchList(expr, names[-nh], parentenv, handlers[-nh]),
names[nh], parentenv, handlers[[nh]])
[[13]]
doTryCatch(return(expr), name, parentenv, handler)
[[14]]
tryCatchList(expr, names[-nh], parentenv, handlers[-nh])
[[15]]
tryCatchOne(expr, names, parentenv, handlers[[1L]])
[[16]]
doTryCatch(return(expr), name, parentenv, handler)
[[17]]
evalq(function (maybe_use_cached = FALSE, trim_tail = 1L)
{
t <- rlang::trace_back()
t <- t[1L:(nrow(t) - trim_tail), ]
t$full_call <- sys.calls()[seq_len(nrow(t))]
if (!maybe_use_cached)
return((.globals$last_r_trace <- t))
ot <- .globals$last_r_trace
if (is.null(ot) || nrow(t) >= nrow(ot) || !identical(t, ot[seq_len(nrow(t)),
])) {
.globals$last_r_trace <- t
}
invisible(.globals$last_r_trace)
}(TRUE, 1L), <environment>)
[[18]]
evalq(function (maybe_use_cached = FALSE, trim_tail = 1L)
{
t <- rlang::trace_back()
t <- t[1L:(nrow(t) - trim_tail), ]
t$full_call <- sys.calls()[seq_len(nrow(t))]
if (!maybe_use_cached)
return((.globals$last_r_trace <- t))
ot <- .globals$last_r_trace
if (is.null(ot) || nrow(t) >= nrow(ot) || !identical(t, ot[seq_len(nrow(t)),
])) {
.globals$last_r_trace <- t
}
invisible(.globals$last_r_trace)
}(TRUE, 1L), <environment>)t-kalinowski commented
Hi, I'm happy to help, but I would need a smaller MRE (minimal reproducible example). Can you please extract out the relevant code that generates the error from that project and place in this thread? Ideally it would be something I can copy+paste+run to see the same error.
How was tensorflow installed? Did you use install_tensorflow()?
Waselia commented
Hi @t-kalinowski, thank you soooo much for your help. I don't know where exactly the code that generates the error, since it is a long function (around 1000 lines or so), and I just ran from the exact original code (I didn't change anything, except for install.packages() and library(), so I think I might just send you the whole chunk, including the generating function part, and the part where I received the error.
Also, I did use the install_tensorflow(), and I also tried install.packages("tensorflow"). I also have the latest version of Python on my computer.
Here's the code for generating the model function:
Code
library(keras)
library(plyr)
library(reticulate)
#py_install("h5py")
#py_install("tensorflow")
library(tensorflow)
##############################################################################################
### Definition of the function neural_network_train_embedding ###
### This function is used to train the neural network and to extract the embedding weights ###
##############################################################################################
### Define the function neural_network_train_embedding
### We have the following inputs:
### seed1 = set the seed for reproducibility
### neurons = number of neurons in the hidden layers of the neural network
### dropout_rates = dropout rates for the past payment information
### starting_values = starting values for the neural network parameters
### network_weights = weights of the individual parts of the loss function
neural_network_train_embedding <- function(seed1, neurons, dropout_rates, starting_values, network_weights){
### Seed
# use_session_with_seed(seed1)
set_random_seed(seed1)
### Dropout
dropout_1_1 <- layer_dropout(rate = dropout_rates[1])
dropout_1_2 <- layer_dropout(rate = dropout_rates[1])
dropout_1_3 <- layer_dropout(rate = dropout_rates[1])
dropout_1_4 <- layer_dropout(rate = dropout_rates[1])
dropout_1_5 <- layer_dropout(rate = dropout_rates[1])
dropout_1_6 <- layer_dropout(rate = dropout_rates[1])
dropout_1_7 <- layer_dropout(rate = dropout_rates[1])
dropout_1_8 <- layer_dropout(rate = dropout_rates[1])
dropout_1_9 <- layer_dropout(rate = dropout_rates[1])
dropout_1_10 <- layer_dropout(rate = dropout_rates[1])
dropout_2_1 <- layer_dropout(rate = dropout_rates[2])
dropout_2_2 <- layer_dropout(rate = dropout_rates[2])
dropout_2_3 <- layer_dropout(rate = dropout_rates[2])
dropout_2_4 <- layer_dropout(rate = dropout_rates[2])
dropout_2_5 <- layer_dropout(rate = dropout_rates[2])
dropout_2_6 <- layer_dropout(rate = dropout_rates[2])
dropout_2_7 <- layer_dropout(rate = dropout_rates[2])
dropout_2_8 <- layer_dropout(rate = dropout_rates[2])
dropout_2_9 <- layer_dropout(rate = dropout_rates[2])
dropout_3_1 <- layer_dropout(rate = dropout_rates[3])
dropout_3_2 <- layer_dropout(rate = dropout_rates[3])
dropout_3_3 <- layer_dropout(rate = dropout_rates[3])
dropout_3_4 <- layer_dropout(rate = dropout_rates[3])
dropout_3_5 <- layer_dropout(rate = dropout_rates[3])
dropout_3_6 <- layer_dropout(rate = dropout_rates[3])
dropout_3_7 <- layer_dropout(rate = dropout_rates[3])
dropout_3_8 <- layer_dropout(rate = dropout_rates[3])
dropout_4_1 <- layer_dropout(rate = dropout_rates[4])
dropout_4_2 <- layer_dropout(rate = dropout_rates[4])
dropout_4_3 <- layer_dropout(rate = dropout_rates[4])
dropout_4_4 <- layer_dropout(rate = dropout_rates[4])
dropout_4_5 <- layer_dropout(rate = dropout_rates[4])
dropout_4_6 <- layer_dropout(rate = dropout_rates[4])
dropout_4_7 <- layer_dropout(rate = dropout_rates[4])
dropout_5_1 <- layer_dropout(rate = dropout_rates[5])
dropout_5_2 <- layer_dropout(rate = dropout_rates[5])
dropout_5_3 <- layer_dropout(rate = dropout_rates[5])
dropout_5_4 <- layer_dropout(rate = dropout_rates[5])
dropout_5_5 <- layer_dropout(rate = dropout_rates[5])
dropout_5_6 <- layer_dropout(rate = dropout_rates[5])
dropout_6_1 <- layer_dropout(rate = dropout_rates[6])
dropout_6_2 <- layer_dropout(rate = dropout_rates[6])
dropout_6_3 <- layer_dropout(rate = dropout_rates[6])
dropout_6_4 <- layer_dropout(rate = dropout_rates[6])
dropout_6_5 <- layer_dropout(rate = dropout_rates[6])
dropout_7_1 <- layer_dropout(rate = dropout_rates[7])
dropout_7_2 <- layer_dropout(rate = dropout_rates[7])
dropout_7_3 <- layer_dropout(rate = dropout_rates[7])
dropout_7_4 <- layer_dropout(rate = dropout_rates[7])
dropout_8_1 <- layer_dropout(rate = dropout_rates[8])
dropout_8_2 <- layer_dropout(rate = dropout_rates[8])
dropout_8_3 <- layer_dropout(rate = dropout_rates[8])
dropout_9_1 <- layer_dropout(rate = dropout_rates[9])
dropout_9_2 <- layer_dropout(rate = dropout_rates[9])
dropout_10_1 <- layer_dropout(rate = dropout_rates[10])
### Inputs
dropout <- layer_input(shape = c(1), dtype = 'int32', name = 'dropout')
cc <- layer_input(shape = c(1), dtype = 'int32', name = 'cc')
AY <- layer_input(shape = c(1), dtype = 'int32', name = 'AY')
AQ <- layer_input(shape = c(1), dtype = 'int32', name = 'AQ')
age <- layer_input(shape = c(1), dtype = 'int32', name = 'age')
inj_part <- layer_input(shape = c(1), dtype = 'int32', name = 'inj_part')
RepDel <- layer_input(shape = c(1), dtype = 'int32', name = 'RepDel')
Time_Known01 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known01')
Time_Known02 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known02')
Time_Known03 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known03')
Time_Known04 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known04')
Time_Known05 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known05')
Time_Known06 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known06')
Time_Known07 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known07')
Time_Known08 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known08')
Time_Known09 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known09')
Time_Known10 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known10')
Time_Known11 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Known11')
Pay00Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay00Info')
Pay01Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay01Info')
Pay02Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay02Info')
Pay03Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay03Info')
Pay04Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay04Info')
Pay05Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay05Info')
Pay06Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay06Info')
Pay07Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay07Info')
Pay08Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay08Info')
Pay09Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay09Info')
Pay10Info <- layer_input(shape = c(1), dtype = 'int32', name = 'Pay10Info')
Time_Predict_Indicator00 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator00')
Time_Predict_Indicator01 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator01')
Time_Predict_Indicator02 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator02')
Time_Predict_Indicator03 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator03')
Time_Predict_Indicator04 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator04')
Time_Predict_Indicator05 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator05')
Time_Predict_Indicator06 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator06')
Time_Predict_Indicator07 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator07')
Time_Predict_Indicator08 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator08')
Time_Predict_Indicator09 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator09')
Time_Predict_Indicator10 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator10')
Time_Predict_Indicator11 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator11')
Time_Predict_Payment00 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment00')
Time_Predict_Payment01 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment01')
Time_Predict_Payment02 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment02')
Time_Predict_Payment03 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment03')
Time_Predict_Payment04 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment04')
Time_Predict_Payment05 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment05')
Time_Predict_Payment06 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment06')
Time_Predict_Payment07 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment07')
Time_Predict_Payment08 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment08')
Time_Predict_Payment09 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment09')
Time_Predict_Payment10 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment10')
Time_Predict_Payment11 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Payment11')
### Embeddings
dropout_embed <- dropout %>%
layer_embedding(input_dim = 1, output_dim = 1, trainable=FALSE, input_length = 1,
weights=list(matrix(c(1)))) %>%
layer_flatten
cc_embed_input <- cc %>%
layer_embedding(input_dim = 51, output_dim = 1, input_length = 1) %>%
layer_flatten
cc_embed_output <- cc %>%
layer_embedding(input_dim = 51, output_dim = 1, input_length = 1) %>%
layer_flatten
AY_embed_input <- AY %>%
layer_embedding(input_dim = 12, output_dim = 1, input_length = 1) %>%
layer_flatten
AY_embed_output <- AY %>%
layer_embedding(input_dim = 12, output_dim = 1, input_length = 1) %>%
layer_flatten
AQ_embed_input <- AQ %>%
layer_embedding(input_dim = 4, output_dim = 1, input_length = 1) %>%
layer_flatten
AQ_embed_output <- AQ %>%
layer_embedding(input_dim = 4, output_dim = 1, input_length = 1) %>%
layer_flatten
age_embed_input <- age %>%
layer_embedding(input_dim = 12, output_dim = 1, input_length = 1) %>%
layer_flatten
age_embed_output <- age %>%
layer_embedding(input_dim = 12, output_dim = 1, input_length = 1) %>%
layer_flatten
inj_part_embed_input <- inj_part %>%
layer_embedding(input_dim = 46, output_dim = 1, input_length = 1) %>%
layer_flatten
inj_part_embed_output <- inj_part %>%
layer_embedding(input_dim = 46, output_dim = 1, input_length = 1) %>%
layer_flatten
RepDel_embed_input <- RepDel %>%
layer_embedding(input_dim = 3, output_dim = 1, input_length = 1) %>%
layer_flatten
RepDel_embed_output <- RepDel %>%
layer_embedding(input_dim = 3, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay00Info_embed_input <- Pay00Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay00Info_embed_output <- Pay00Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay01Info_embed_input <- Pay01Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay01Info_embed_output <- Pay01Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay02Info_embed_input <- Pay02Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay02Info_embed_output <- Pay02Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay03Info_embed_input <- Pay03Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay03Info_embed_output <- Pay03Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay04Info_embed_input <- Pay04Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay04Info_embed_output <- Pay04Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay05Info_embed_input <- Pay05Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay05Info_embed_output <- Pay05Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay06Info_embed_input <- Pay06Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay06Info_embed_output <- Pay06Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay07Info_embed_input <- Pay07Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay07Info_embed_output <- Pay07Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay08Info_embed_input <- Pay08Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay08Info_embed_output <- Pay08Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay09Info_embed_input <- Pay09Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay09Info_embed_output <- Pay09Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay10Info_embed_input <- Pay10Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
Pay10Info_embed_output <- Pay10Info %>%
layer_embedding(input_dim = 6, output_dim = 1, input_length = 1) %>%
layer_flatten
### Time 00 - Time 11
features_input <- list(cc_embed_input, AQ_embed_input, age_embed_input, inj_part_embed_input, RepDel_embed_input) %>%
layer_concatenate
AY_embed_input_NN <- AY_embed_input %>%
layer_dense(units = neurons[1], activation='linear')
features_output <- list(cc_embed_output, AQ_embed_output, age_embed_output, inj_part_embed_output, RepDel_embed_output) %>%
layer_concatenate
AY_output_indicator <- AY_embed_output %>%
layer_dense(units = 1, activation='linear', name="AY_output_indicator",
weights = list(array(0,dim=c(1,1)),array(0,dim=c(1))))
AY_output_mean <- AY_embed_output %>%
layer_dense(units = 1, activation='linear', name="AY_output_mean",
weights = list(array(0,dim=c(1,1)),array(0,dim=c(1))))
### Time 00
hidden_0_pre <- features_input %>%
layer_dense(units=neurons[1], activation='linear')
hidden_0 <- list(hidden_0_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator00_pre2 <- list(payment_indicator00_pre, features_output) %>%
layer_concatenate(name = "pay_ind00") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(log(starting_values[1,1]/(1-starting_values[1,1])),dim=c(1))))
payment_indicator00 <- list(payment_indicator00_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean00_pre2 <- list(payment_mean00_pre, features_output) %>%
layer_concatenate(name = "pay_mean00") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(starting_values[1,2],dim=c(1))))
payment_mean00 <- list(payment_mean00_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 01
hidden_1_pre <- list(features_input, Pay00Info_embed_input) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_1 <- list(hidden_1_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator01_pre <- hidden_1 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator01_pre2 <- list(payment_indicator01_pre, features_output, Pay00Info_embed_output) %>%
layer_concatenate(name = "pay_ind01") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+6),dim=c(neurons[3]+6,1)),array(log(starting_values[2,1]/(1-starting_values[2,1])),dim=c(1))))
payment_indicator01 <- list(payment_indicator01_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean01_pre <- hidden_1 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean01_pre2 <- list(payment_mean01_pre, features_output, Pay00Info_embed_output) %>%
layer_concatenate(name = "pay_mean01") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+6),dim=c(neurons[3]+6,1)),array(starting_values[2,2],dim=c(1))))
payment_mean01 <- list(payment_mean01_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 02
dropout_1_10_applied <- dropout_embed %>%
dropout_1_10(training = TRUE)
Pay01Info_embed_input_dropout_02 <- list(Pay01Info_embed_input, dropout_1_10_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_02 <- list(Pay01Info_embed_output, dropout_1_10_applied, Time_Known01) %>%
layer_multiply()
hidden_2_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_02) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_2 <- list(hidden_2_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator02_pre <- hidden_2 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator02_pre2 <- list(payment_indicator02_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_02) %>%
layer_concatenate(name = "pay_ind02") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+7),dim=c(neurons[3]+7,1)),array(log(starting_values[3,1]/(1-starting_values[3,1])),dim=c(1))))
payment_indicator02 <- list(payment_indicator02_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean02_pre <- hidden_2 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean02_pre2 <- list(payment_mean02_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_02) %>%
layer_concatenate(name = "pay_mean02") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+7),dim=c(neurons[3]+7,1)),array(starting_values[3,2],dim=c(1))))
payment_mean02 <- list(payment_mean02_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 03
dropout_1_9_applied <- dropout_embed %>%
dropout_1_9(training = TRUE)
dropout_2_9_applied <- dropout_embed %>%
dropout_2_9(training = TRUE)
Pay01Info_embed_input_dropout_03 <- list(Pay01Info_embed_input, dropout_2_9_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_03 <- list(Pay01Info_embed_output, dropout_2_9_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_03 <- list(Pay02Info_embed_input, dropout_2_9_applied, dropout_1_9_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_03 <- list(Pay02Info_embed_output, dropout_2_9_applied, dropout_1_9_applied, Time_Known02) %>%
layer_multiply()
hidden_3_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_03, Pay02Info_embed_input_dropout_03) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_3 <- list(hidden_3_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator03_pre <- hidden_3 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator03_pre2 <- list(payment_indicator03_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_03, Pay02Info_embed_output_dropout_03) %>%
layer_concatenate(name = "pay_ind03") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+8),dim=c(neurons[3]+8,1)),array(log(starting_values[4,1]/(1-starting_values[4,1])),dim=c(1))))
payment_indicator03 <- list(payment_indicator03_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean03_pre <- hidden_3 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean03_pre2 <- list(payment_mean03_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_03, Pay02Info_embed_output_dropout_03) %>%
layer_concatenate(name = "pay_mean03") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+8),dim=c(neurons[3]+8,1)),array(starting_values[4,2],dim=c(1))))
payment_mean03 <- list(payment_mean03_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 04
dropout_1_8_applied <- dropout_embed %>%
dropout_1_8(training = TRUE)
dropout_2_8_applied <- dropout_embed %>%
dropout_2_8(training = TRUE)
dropout_3_8_applied <- dropout_embed %>%
dropout_3_8(training = TRUE)
Pay01Info_embed_input_dropout_04 <- list(Pay01Info_embed_input, dropout_3_8_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_04 <- list(Pay01Info_embed_output, dropout_3_8_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_04 <- list(Pay02Info_embed_input, dropout_3_8_applied, dropout_2_8_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_04 <- list(Pay02Info_embed_output, dropout_3_8_applied, dropout_2_8_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_04 <- list(Pay03Info_embed_input, dropout_3_8_applied, dropout_2_8_applied, dropout_1_8_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_04 <- list(Pay03Info_embed_output, dropout_3_8_applied, dropout_2_8_applied, dropout_1_8_applied, Time_Known03) %>%
layer_multiply()
hidden_4_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_04, Pay02Info_embed_input_dropout_04, Pay03Info_embed_input_dropout_04) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_4 <- list(hidden_4_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator04_pre <- hidden_4 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator04_pre2 <- list(payment_indicator04_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_04, Pay02Info_embed_output_dropout_04, Pay03Info_embed_output_dropout_04) %>%
layer_concatenate(name = "pay_ind04") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+9),dim=c(neurons[3]+9,1)),array(log(starting_values[5,1]/(1-starting_values[5,1])),dim=c(1))))
payment_indicator04 <- list(payment_indicator04_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean04_pre <- hidden_4 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean04_pre2 <- list(payment_mean04_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_04, Pay02Info_embed_output_dropout_04, Pay03Info_embed_output_dropout_04) %>%
layer_concatenate(name = "pay_mean04") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+9),dim=c(neurons[3]+9,1)),array(starting_values[5,2],dim=c(1))))
payment_mean04 <- list(payment_mean04_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 05
dropout_1_7_applied <- dropout_embed %>%
dropout_1_7(training = TRUE)
dropout_2_7_applied <- dropout_embed %>%
dropout_2_7(training = TRUE)
dropout_3_7_applied <- dropout_embed %>%
dropout_3_7(training = TRUE)
dropout_4_7_applied <- dropout_embed %>%
dropout_4_7(training = TRUE)
Pay01Info_embed_input_dropout_05 <- list(Pay01Info_embed_input, dropout_4_7_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_05 <- list(Pay01Info_embed_output, dropout_4_7_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_05 <- list(Pay02Info_embed_input, dropout_4_7_applied, dropout_3_7_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_05 <- list(Pay02Info_embed_output, dropout_4_7_applied, dropout_3_7_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_05 <- list(Pay03Info_embed_input, dropout_4_7_applied, dropout_3_7_applied, dropout_2_7_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_05 <- list(Pay03Info_embed_output, dropout_4_7_applied, dropout_3_7_applied, dropout_2_7_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_05 <- list(Pay04Info_embed_input, dropout_4_7_applied, dropout_3_7_applied, dropout_2_7_applied, dropout_1_7_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_05 <- list(Pay04Info_embed_output, dropout_4_7_applied, dropout_3_7_applied, dropout_2_7_applied, dropout_1_7_applied, Time_Known04) %>%
layer_multiply()
hidden_5_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_05, Pay02Info_embed_input_dropout_05, Pay03Info_embed_input_dropout_05, Pay04Info_embed_input_dropout_05) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_5 <- list(hidden_5_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator05_pre <- hidden_5 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator05_pre2 <- list(payment_indicator05_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_05, Pay02Info_embed_output_dropout_05, Pay03Info_embed_output_dropout_05, Pay04Info_embed_output_dropout_05) %>%
layer_concatenate(name = "pay_ind05") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+10),dim=c(neurons[3]+10,1)),array(log(starting_values[6,1]/(1-starting_values[6,1])),dim=c(1))))
payment_indicator05 <- list(payment_indicator05_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean05_pre <- hidden_5 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean05_pre2 <- list(payment_mean05_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_05, Pay02Info_embed_output_dropout_05, Pay03Info_embed_output_dropout_05, Pay04Info_embed_output_dropout_05) %>%
layer_concatenate(name = "pay_mean05") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+10),dim=c(neurons[3]+10,1)),array(starting_values[6,2],dim=c(1))))
payment_mean05 <- list(payment_mean05_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 06
dropout_1_6_applied <- dropout_embed %>%
dropout_1_6(training = TRUE)
dropout_2_6_applied <- dropout_embed %>%
dropout_2_6(training = TRUE)
dropout_3_6_applied <- dropout_embed %>%
dropout_3_6(training = TRUE)
dropout_4_6_applied <- dropout_embed %>%
dropout_4_6(training = TRUE)
dropout_5_6_applied <- dropout_embed %>%
dropout_5_6(training = TRUE)
Pay01Info_embed_input_dropout_06 <- list(Pay01Info_embed_input, dropout_5_6_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_06 <- list(Pay01Info_embed_output, dropout_5_6_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_06 <- list(Pay02Info_embed_input, dropout_5_6_applied, dropout_4_6_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_06 <- list(Pay02Info_embed_output, dropout_5_6_applied, dropout_4_6_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_06 <- list(Pay03Info_embed_input, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_06 <- list(Pay03Info_embed_output, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_06 <- list(Pay04Info_embed_input, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, dropout_2_6_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_06 <- list(Pay04Info_embed_output, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, dropout_2_6_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_06 <- list(Pay05Info_embed_input, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, dropout_2_6_applied, dropout_1_6_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_06 <- list(Pay05Info_embed_output, dropout_5_6_applied, dropout_4_6_applied, dropout_3_6_applied, dropout_2_6_applied, dropout_1_6_applied, Time_Known05) %>%
layer_multiply()
hidden_6_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_06, Pay02Info_embed_input_dropout_06, Pay03Info_embed_input_dropout_06, Pay04Info_embed_input_dropout_06, Pay05Info_embed_input_dropout_06) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_6 <- list(hidden_6_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator06_pre <- hidden_6 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator06_pre2 <- list(payment_indicator06_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_06, Pay02Info_embed_output_dropout_06, Pay03Info_embed_output_dropout_06, Pay04Info_embed_output_dropout_06, Pay05Info_embed_output_dropout_06) %>%
layer_concatenate(name = "pay_ind06") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+11),dim=c(neurons[3]+11,1)),array(log(starting_values[7,1]/(1-starting_values[7,1])),dim=c(1))))
payment_indicator06 <- list(payment_indicator06_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean06_pre <- hidden_6 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean06_pre2 <- list(payment_mean06_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_06, Pay02Info_embed_output_dropout_06, Pay03Info_embed_output_dropout_06, Pay04Info_embed_output_dropout_06, Pay05Info_embed_output_dropout_06) %>%
layer_concatenate(name = "pay_mean06") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+11),dim=c(neurons[3]+11,1)),array(starting_values[7,2],dim=c(1))))
payment_mean06 <- list(payment_mean06_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 07
dropout_1_5_applied <- dropout_embed %>%
dropout_1_5(training = TRUE)
dropout_2_5_applied <- dropout_embed %>%
dropout_2_5(training = TRUE)
dropout_3_5_applied <- dropout_embed %>%
dropout_3_5(training = TRUE)
dropout_4_5_applied <- dropout_embed %>%
dropout_4_5(training = TRUE)
dropout_5_5_applied <- dropout_embed %>%
dropout_5_5(training = TRUE)
dropout_6_5_applied <- dropout_embed %>%
dropout_6_5(training = TRUE)
Pay01Info_embed_input_dropout_07 <- list(Pay01Info_embed_input, dropout_6_5_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_07 <- list(Pay01Info_embed_output, dropout_6_5_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_07 <- list(Pay02Info_embed_input, dropout_6_5_applied, dropout_5_5_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_07 <- list(Pay02Info_embed_output, dropout_6_5_applied, dropout_5_5_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_07 <- list(Pay03Info_embed_input, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_07 <- list(Pay03Info_embed_output, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_07 <- list(Pay04Info_embed_input, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_07 <- list(Pay04Info_embed_output, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_07 <- list(Pay05Info_embed_input, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, dropout_2_5_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_07 <- list(Pay05Info_embed_output, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, dropout_2_5_applied, Time_Known05) %>%
layer_multiply()
Pay06Info_embed_input_dropout_07 <- list(Pay06Info_embed_input, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, dropout_2_5_applied, dropout_1_5_applied, Time_Known06) %>%
layer_multiply()
Pay06Info_embed_output_dropout_07 <- list(Pay06Info_embed_output, dropout_6_5_applied, dropout_5_5_applied, dropout_4_5_applied, dropout_3_5_applied, dropout_2_5_applied, dropout_1_5_applied, Time_Known06) %>%
layer_multiply()
hidden_7_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_07, Pay02Info_embed_input_dropout_07, Pay03Info_embed_input_dropout_07, Pay04Info_embed_input_dropout_07, Pay05Info_embed_input_dropout_07, Pay06Info_embed_input_dropout_07) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_7 <- list(hidden_7_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator07_pre <- hidden_7 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator07_pre2 <- list(payment_indicator07_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_07, Pay02Info_embed_output_dropout_07, Pay03Info_embed_output_dropout_07, Pay04Info_embed_output_dropout_07, Pay05Info_embed_output_dropout_07, Pay06Info_embed_output_dropout_07) %>%
layer_concatenate(name = "pay_ind07") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+12),dim=c(neurons[3]+12,1)),array(log(starting_values[8,1]/(1-starting_values[8,1])),dim=c(1))))
payment_indicator07 <- list(payment_indicator07_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean07_pre <- hidden_7 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean07_pre2 <- list(payment_mean07_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_07, Pay02Info_embed_output_dropout_07, Pay03Info_embed_output_dropout_07, Pay04Info_embed_output_dropout_07, Pay05Info_embed_output_dropout_07, Pay06Info_embed_output_dropout_07) %>%
layer_concatenate(name = "pay_mean07") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+12),dim=c(neurons[3]+12,1)),array(starting_values[8,2],dim=c(1))))
payment_mean07 <- list(payment_mean07_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 08
dropout_1_4_applied <- dropout_embed %>%
dropout_1_4(training = TRUE)
dropout_2_4_applied <- dropout_embed %>%
dropout_2_4(training = TRUE)
dropout_3_4_applied <- dropout_embed %>%
dropout_3_4(training = TRUE)
dropout_4_4_applied <- dropout_embed %>%
dropout_4_4(training = TRUE)
dropout_5_4_applied <- dropout_embed %>%
dropout_5_4(training = TRUE)
dropout_6_4_applied <- dropout_embed %>%
dropout_6_4(training = TRUE)
dropout_7_4_applied <- dropout_embed %>%
dropout_7_4(training = TRUE)
Pay01Info_embed_input_dropout_08 <- list(Pay01Info_embed_input, dropout_7_4_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_08 <- list(Pay01Info_embed_output, dropout_7_4_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_08 <- list(Pay02Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_08 <- list(Pay02Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_08 <- list(Pay03Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_08 <- list(Pay03Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_08 <- list(Pay04Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_08 <- list(Pay04Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_08 <- list(Pay05Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_08 <- list(Pay05Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, Time_Known05) %>%
layer_multiply()
Pay06Info_embed_input_dropout_08 <- list(Pay06Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, dropout_2_4_applied, Time_Known06) %>%
layer_multiply()
Pay06Info_embed_output_dropout_08 <- list(Pay06Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, dropout_2_4_applied, Time_Known06) %>%
layer_multiply()
Pay07Info_embed_input_dropout_08 <- list(Pay07Info_embed_input, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, dropout_2_4_applied, dropout_1_4_applied, Time_Known07) %>%
layer_multiply()
Pay07Info_embed_output_dropout_08 <- list(Pay07Info_embed_output, dropout_7_4_applied, dropout_6_4_applied, dropout_5_4_applied, dropout_4_4_applied, dropout_3_4_applied, dropout_2_4_applied, dropout_1_4_applied, Time_Known07) %>%
layer_multiply()
hidden_8_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_08, Pay02Info_embed_input_dropout_08, Pay03Info_embed_input_dropout_08, Pay04Info_embed_input_dropout_08, Pay05Info_embed_input_dropout_08, Pay06Info_embed_input_dropout_08, Pay07Info_embed_input_dropout_08) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_8 <- list(hidden_8_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator08_pre <- hidden_8 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator08_pre2 <- list(payment_indicator08_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_08, Pay02Info_embed_output_dropout_08, Pay03Info_embed_output_dropout_08, Pay04Info_embed_output_dropout_08, Pay05Info_embed_output_dropout_08, Pay06Info_embed_output_dropout_08, Pay07Info_embed_output_dropout_08) %>%
layer_concatenate(name = "pay_ind08") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+13),dim=c(neurons[3]+13,1)),array(log(starting_values[9,1]/(1-starting_values[9,1])),dim=c(1))))
payment_indicator08 <- list(payment_indicator08_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean08_pre <- hidden_8 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean08_pre2 <- list(payment_mean08_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_08, Pay02Info_embed_output_dropout_08, Pay03Info_embed_output_dropout_08, Pay04Info_embed_output_dropout_08, Pay05Info_embed_output_dropout_08, Pay06Info_embed_output_dropout_08, Pay07Info_embed_output_dropout_08) %>%
layer_concatenate(name = "pay_mean08") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+13),dim=c(neurons[3]+13,1)),array(starting_values[9,2],dim=c(1))))
payment_mean08 <- list(payment_mean08_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 09
dropout_1_3_applied <- dropout_embed %>%
dropout_1_3(training = TRUE)
dropout_2_3_applied <- dropout_embed %>%
dropout_2_3(training = TRUE)
dropout_3_3_applied <- dropout_embed %>%
dropout_3_3(training = TRUE)
dropout_4_3_applied <- dropout_embed %>%
dropout_4_3(training = TRUE)
dropout_5_3_applied <- dropout_embed %>%
dropout_5_3(training = TRUE)
dropout_6_3_applied <- dropout_embed %>%
dropout_6_3(training = TRUE)
dropout_7_3_applied <- dropout_embed %>%
dropout_7_3(training = TRUE)
dropout_8_3_applied <- dropout_embed %>%
dropout_8_3(training = TRUE)
Pay01Info_embed_input_dropout_09 <- list(Pay01Info_embed_input, dropout_8_3_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_09 <- list(Pay01Info_embed_output, dropout_8_3_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_09 <- list(Pay02Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_09 <- list(Pay02Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_09 <- list(Pay03Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_09 <- list(Pay03Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_09 <- list(Pay04Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_09 <- list(Pay04Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_09 <- list(Pay05Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_09 <- list(Pay05Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, Time_Known05) %>%
layer_multiply()
Pay06Info_embed_input_dropout_09 <- list(Pay06Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, Time_Known06) %>%
layer_multiply()
Pay06Info_embed_output_dropout_09 <- list(Pay06Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, Time_Known06) %>%
layer_multiply()
Pay07Info_embed_input_dropout_09 <- list(Pay07Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, dropout_2_3_applied, Time_Known07) %>%
layer_multiply()
Pay07Info_embed_output_dropout_09 <- list(Pay07Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, dropout_2_3_applied, Time_Known07) %>%
layer_multiply()
Pay08Info_embed_input_dropout_09 <- list(Pay08Info_embed_input, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, dropout_2_3_applied, dropout_1_3_applied, Time_Known08) %>%
layer_multiply()
Pay08Info_embed_output_dropout_09 <- list(Pay08Info_embed_output, dropout_8_3_applied, dropout_7_3_applied, dropout_6_3_applied, dropout_5_3_applied, dropout_4_3_applied, dropout_3_3_applied, dropout_2_3_applied, dropout_1_3_applied, Time_Known08) %>%
layer_multiply()
hidden_9_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_09, Pay02Info_embed_input_dropout_09, Pay03Info_embed_input_dropout_09, Pay04Info_embed_input_dropout_09, Pay05Info_embed_input_dropout_09, Pay06Info_embed_input_dropout_09, Pay07Info_embed_input_dropout_09, Pay08Info_embed_input_dropout_09) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_9 <- list(hidden_9_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator09_pre <- hidden_9 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator09_pre2 <- list(payment_indicator09_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_09, Pay02Info_embed_output_dropout_09, Pay03Info_embed_output_dropout_09, Pay04Info_embed_output_dropout_09, Pay05Info_embed_output_dropout_09, Pay06Info_embed_output_dropout_09, Pay07Info_embed_output_dropout_09, Pay08Info_embed_output_dropout_09) %>%
layer_concatenate(name = "pay_ind09") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+14),dim=c(neurons[3]+14,1)),array(log(starting_values[10,1]/(1-starting_values[10,1])),dim=c(1))))
payment_indicator09 <- list(payment_indicator09_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean09_pre <- hidden_9 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean09_pre2 <- list(payment_mean09_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_09, Pay02Info_embed_output_dropout_09, Pay03Info_embed_output_dropout_09, Pay04Info_embed_output_dropout_09, Pay05Info_embed_output_dropout_09, Pay06Info_embed_output_dropout_09, Pay07Info_embed_output_dropout_09, Pay08Info_embed_output_dropout_09) %>%
layer_concatenate(name = "pay_mean09") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+14),dim=c(neurons[3]+14,1)),array(starting_values[10,2],dim=c(1))))
payment_mean09 <- list(payment_mean09_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 10
dropout_1_2_applied <- dropout_embed %>%
dropout_1_2(training = TRUE)
dropout_2_2_applied <- dropout_embed %>%
dropout_2_2(training = TRUE)
dropout_3_2_applied <- dropout_embed %>%
dropout_3_2(training = TRUE)
dropout_4_2_applied <- dropout_embed %>%
dropout_4_2(training = TRUE)
dropout_5_2_applied <- dropout_embed %>%
dropout_5_2(training = TRUE)
dropout_6_2_applied <- dropout_embed %>%
dropout_6_2(training = TRUE)
dropout_7_2_applied <- dropout_embed %>%
dropout_7_2(training = TRUE)
dropout_8_2_applied <- dropout_embed %>%
dropout_8_2(training = TRUE)
dropout_9_2_applied <- dropout_embed %>%
dropout_9_2(training = TRUE)
Pay01Info_embed_input_dropout_10 <- list(Pay01Info_embed_input, dropout_9_2_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_10 <- list(Pay01Info_embed_output, dropout_9_2_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_10 <- list(Pay02Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_10 <- list(Pay02Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_10 <- list(Pay03Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_10 <- list(Pay03Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_10 <- list(Pay04Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_10 <- list(Pay04Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_10 <- list(Pay05Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_10 <- list(Pay05Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, Time_Known05) %>%
layer_multiply()
Pay06Info_embed_input_dropout_10 <- list(Pay06Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, Time_Known06) %>%
layer_multiply()
Pay06Info_embed_output_dropout_10 <- list(Pay06Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, Time_Known06) %>%
layer_multiply()
Pay07Info_embed_input_dropout_10 <- list(Pay07Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, Time_Known07) %>%
layer_multiply()
Pay07Info_embed_output_dropout_10 <- list(Pay07Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, Time_Known07) %>%
layer_multiply()
Pay08Info_embed_input_dropout_10 <- list(Pay08Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, dropout_2_2_applied, Time_Known08) %>%
layer_multiply()
Pay08Info_embed_output_dropout_10 <- list(Pay08Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, dropout_2_2_applied, Time_Known08) %>%
layer_multiply()
Pay09Info_embed_input_dropout_10 <- list(Pay09Info_embed_input, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, dropout_2_2_applied, dropout_1_2_applied, Time_Known09) %>%
layer_multiply()
Pay09Info_embed_output_dropout_10 <- list(Pay09Info_embed_output, dropout_9_2_applied, dropout_8_2_applied, dropout_7_2_applied, dropout_6_2_applied, dropout_5_2_applied, dropout_4_2_applied, dropout_3_2_applied, dropout_2_2_applied, dropout_1_2_applied, Time_Known09) %>%
layer_multiply()
hidden_10_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_10, Pay02Info_embed_input_dropout_10, Pay03Info_embed_input_dropout_10, Pay04Info_embed_input_dropout_10, Pay05Info_embed_input_dropout_10, Pay06Info_embed_input_dropout_10, Pay07Info_embed_input_dropout_10, Pay08Info_embed_input_dropout_10, Pay09Info_embed_input_dropout_10) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_10 <- list(hidden_10_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator10_pre <- hidden_10 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator10_pre2 <- list(payment_indicator10_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_10, Pay02Info_embed_output_dropout_10, Pay03Info_embed_output_dropout_10, Pay04Info_embed_output_dropout_10, Pay05Info_embed_output_dropout_10, Pay06Info_embed_output_dropout_10, Pay07Info_embed_output_dropout_10, Pay08Info_embed_output_dropout_10, Pay09Info_embed_output_dropout_10) %>%
layer_concatenate(name = "pay_ind10") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+15),dim=c(neurons[3]+15,1)),array(log(starting_values[11,1]/(1-starting_values[11,1])),dim=c(1))))
payment_indicator10 <- list(payment_indicator10_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean10_pre <- hidden_10 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean10_pre2 <- list(payment_mean10_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_10, Pay02Info_embed_output_dropout_10, Pay03Info_embed_output_dropout_10, Pay04Info_embed_output_dropout_10, Pay05Info_embed_output_dropout_10, Pay06Info_embed_output_dropout_10, Pay07Info_embed_output_dropout_10, Pay08Info_embed_output_dropout_10, Pay09Info_embed_output_dropout_10) %>%
layer_concatenate(name = "pay_mean10") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+15),dim=c(neurons[3]+15,1)),array(starting_values[11,2],dim=c(1))))
payment_mean10 <- list(payment_mean10_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Time 11
dropout_1_1_applied <- dropout_embed %>%
dropout_1_1(training = TRUE)
dropout_2_1_applied <- dropout_embed %>%
dropout_2_1(training = TRUE)
dropout_3_1_applied <- dropout_embed %>%
dropout_3_1(training = TRUE)
dropout_4_1_applied <- dropout_embed %>%
dropout_4_1(training = TRUE)
dropout_5_1_applied <- dropout_embed %>%
dropout_5_1(training = TRUE)
dropout_6_1_applied <- dropout_embed %>%
dropout_6_1(training = TRUE)
dropout_7_1_applied <- dropout_embed %>%
dropout_7_1(training = TRUE)
dropout_8_1_applied <- dropout_embed %>%
dropout_8_1(training = TRUE)
dropout_9_1_applied <- dropout_embed %>%
dropout_9_1(training = TRUE)
dropout_10_1_applied <- dropout_embed %>%
dropout_10_1(training = TRUE)
Pay01Info_embed_input_dropout_11 <- list(Pay01Info_embed_input, dropout_10_1_applied, Time_Known01) %>%
layer_multiply()
Pay01Info_embed_output_dropout_11 <- list(Pay01Info_embed_output, dropout_10_1_applied, Time_Known01) %>%
layer_multiply()
Pay02Info_embed_input_dropout_11 <- list(Pay02Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, Time_Known02) %>%
layer_multiply()
Pay02Info_embed_output_dropout_11 <- list(Pay02Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, Time_Known02) %>%
layer_multiply()
Pay03Info_embed_input_dropout_11 <- list(Pay03Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, Time_Known03) %>%
layer_multiply()
Pay03Info_embed_output_dropout_11 <- list(Pay03Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, Time_Known03) %>%
layer_multiply()
Pay04Info_embed_input_dropout_11 <- list(Pay04Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, Time_Known04) %>%
layer_multiply()
Pay04Info_embed_output_dropout_11 <- list(Pay04Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, Time_Known04) %>%
layer_multiply()
Pay05Info_embed_input_dropout_11 <- list(Pay05Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, Time_Known05) %>%
layer_multiply()
Pay05Info_embed_output_dropout_11 <- list(Pay05Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, Time_Known05) %>%
layer_multiply()
Pay06Info_embed_input_dropout_11 <- list(Pay06Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, Time_Known06) %>%
layer_multiply()
Pay06Info_embed_output_dropout_11 <- list(Pay06Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, Time_Known06) %>%
layer_multiply()
Pay07Info_embed_input_dropout_11 <- list(Pay07Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, Time_Known07) %>%
layer_multiply()
Pay07Info_embed_output_dropout_11 <- list(Pay07Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, Time_Known07) %>%
layer_multiply()
Pay08Info_embed_input_dropout_11 <- list(Pay08Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, Time_Known08) %>%
layer_multiply()
Pay08Info_embed_output_dropout_11 <- list(Pay08Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, Time_Known08) %>%
layer_multiply()
Pay09Info_embed_input_dropout_11 <- list(Pay09Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, dropout_2_1_applied, Time_Known09) %>%
layer_multiply()
Pay09Info_embed_output_dropout_11 <- list(Pay09Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, dropout_2_1_applied, Time_Known09) %>%
layer_multiply()
Pay10Info_embed_input_dropout_11 <- list(Pay10Info_embed_input, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, dropout_2_1_applied, dropout_1_1_applied, Time_Known10) %>%
layer_multiply()
Pay10Info_embed_output_dropout_11 <- list(Pay10Info_embed_output, dropout_10_1_applied, dropout_9_1_applied, dropout_8_1_applied, dropout_7_1_applied, dropout_6_1_applied, dropout_5_1_applied, dropout_4_1_applied, dropout_3_1_applied, dropout_2_1_applied, dropout_1_1_applied, Time_Known10) %>%
layer_multiply()
hidden_11_pre <- list(features_input, Pay00Info_embed_input, Pay01Info_embed_input_dropout_11, Pay02Info_embed_input_dropout_11, Pay03Info_embed_input_dropout_11, Pay04Info_embed_input_dropout_11, Pay05Info_embed_input_dropout_11, Pay06Info_embed_input_dropout_11, Pay07Info_embed_input_dropout_11, Pay08Info_embed_input_dropout_11, Pay09Info_embed_input_dropout_11, Pay10Info_embed_input_dropout_11) %>%
layer_concatenate() %>%
layer_dense(units=neurons[1], activation='linear')
hidden_11 <- list(hidden_11_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator11_pre <- hidden_11 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator11_pre2 <- list(payment_indicator11_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_11, Pay02Info_embed_output_dropout_11, Pay03Info_embed_output_dropout_11, Pay04Info_embed_output_dropout_11, Pay05Info_embed_output_dropout_11, Pay06Info_embed_output_dropout_11, Pay07Info_embed_output_dropout_11, Pay08Info_embed_output_dropout_11, Pay09Info_embed_output_dropout_11, Pay10Info_embed_output_dropout_11) %>%
layer_concatenate(name = "pay_ind11") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+16),dim=c(neurons[3]+16,1)),array(log(starting_values[12,1]/(1-starting_values[12,1])),dim=c(1))))
payment_indicator11 <- list(payment_indicator11_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
payment_mean11_pre <- hidden_11 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_mean11_pre2 <- list(payment_mean11_pre, features_output, Pay00Info_embed_output, Pay01Info_embed_output_dropout_11, Pay02Info_embed_output_dropout_11, Pay03Info_embed_output_dropout_11, Pay04Info_embed_output_dropout_11, Pay05Info_embed_output_dropout_11, Pay06Info_embed_output_dropout_11, Pay07Info_embed_output_dropout_11, Pay08Info_embed_output_dropout_11, Pay09Info_embed_output_dropout_11, Pay10Info_embed_output_dropout_11) %>%
layer_concatenate(name = "pay_mean11") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+16),dim=c(neurons[3]+16,1)),array(starting_values[12,2],dim=c(1))))
payment_mean11 <- list(payment_mean11_pre2, AY_output_mean) %>%
layer_add %>%
layer_dense(units = 1, activation = 'linear', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
### Multiply the outputs with the Time_Predict indicators (0/1)
### Time 00
payment_indicator00_output <- list(payment_indicator00, Time_Predict_Indicator00) %>%
layer_multiply(name = 'payment_indicator00_output')
payment_mean00_output <- list(payment_mean00, Time_Predict_Payment00) %>%
layer_multiply(name = 'payment_mean00_output')
### Time 01
payment_indicator01_output <- list(payment_indicator01, Time_Predict_Indicator01) %>%
layer_multiply(name = 'payment_indicator01_output')
payment_mean01_output <- list(payment_mean01, Time_Predict_Payment01) %>%
layer_multiply(name = 'payment_mean01_output')
### Time 02
payment_indicator02_output <- list(payment_indicator02, Time_Predict_Indicator02) %>%
layer_multiply(name = 'payment_indicator02_output')
payment_mean02_output <- list(payment_mean02, Time_Predict_Payment02) %>%
layer_multiply(name = 'payment_mean02_output')
### Time 03
payment_indicator03_output <- list(payment_indicator03, Time_Predict_Indicator03) %>%
layer_multiply(name = 'payment_indicator03_output')
payment_mean03_output <- list(payment_mean03, Time_Predict_Payment03) %>%
layer_multiply(name = 'payment_mean03_output')
### Time 04
payment_indicator04_output <- list(payment_indicator04, Time_Predict_Indicator04) %>%
layer_multiply(name = 'payment_indicator04_output')
payment_mean04_output <- list(payment_mean04, Time_Predict_Payment04) %>%
layer_multiply(name = 'payment_mean04_output')
### Time 05
payment_indicator05_output <- list(payment_indicator05, Time_Predict_Indicator05) %>%
layer_multiply(name = 'payment_indicator05_output')
payment_mean05_output <- list(payment_mean05, Time_Predict_Payment05) %>%
layer_multiply(name = 'payment_mean05_output')
### Time 06
payment_indicator06_output <- list(payment_indicator06, Time_Predict_Indicator06) %>%
layer_multiply(name = 'payment_indicator06_output')
payment_mean06_output <- list(payment_mean06, Time_Predict_Payment06) %>%
layer_multiply(name = 'payment_mean06_output')
### Time 07
payment_indicator07_output <- list(payment_indicator07, Time_Predict_Indicator07) %>%
layer_multiply(name = 'payment_indicator07_output')
payment_mean07_output <- list(payment_mean07, Time_Predict_Payment07) %>%
layer_multiply(name = 'payment_mean07_output')
### Time 08
payment_indicator08_output <- list(payment_indicator08, Time_Predict_Indicator08) %>%
layer_multiply(name = 'payment_indicator08_output')
payment_mean08_output <- list(payment_mean08, Time_Predict_Payment08) %>%
layer_multiply(name = 'payment_mean08_output')
### Time 09
payment_indicator09_output <- list(payment_indicator09, Time_Predict_Indicator09) %>%
layer_multiply(name = 'payment_indicator09_output')
payment_mean09_output <- list(payment_mean09, Time_Predict_Payment09) %>%
layer_multiply(name = 'payment_mean09_output')
### Time 10
payment_indicator10_output <- list(payment_indicator10, Time_Predict_Indicator10) %>%
layer_multiply(name = 'payment_indicator10_output')
payment_mean10_output <- list(payment_mean10, Time_Predict_Payment10) %>%
layer_multiply(name = 'payment_mean10_output')
### Time 11
payment_indicator11_output <- list(payment_indicator11, Time_Predict_Indicator11) %>%
layer_multiply(name = 'payment_indicator11_output')
payment_mean11_output <- list(payment_mean11, Time_Predict_Payment11) %>%
layer_multiply(name = 'payment_mean11_output')
### Keras model
model <- keras_model(inputs = c(dropout, cc, AY, AQ, age, inj_part, RepDel,
Time_Known01, Time_Known02, Time_Known03, Time_Known04, Time_Known05, Time_Known06,
Time_Known07, Time_Known08, Time_Known09, Time_Known10, Time_Known11,
Pay00Info, Pay01Info, Pay02Info, Pay03Info, Pay04Info, Pay05Info,
Pay06Info, Pay07Info, Pay08Info, Pay09Info, Pay10Info,
Time_Predict_Indicator00, Time_Predict_Indicator01, Time_Predict_Indicator02, Time_Predict_Indicator03, Time_Predict_Indicator04, Time_Predict_Indicator05,
Time_Predict_Indicator06, Time_Predict_Indicator07, Time_Predict_Indicator08, Time_Predict_Indicator09, Time_Predict_Indicator10, Time_Predict_Indicator11,
Time_Predict_Payment00, Time_Predict_Payment01, Time_Predict_Payment02, Time_Predict_Payment03, Time_Predict_Payment04, Time_Predict_Payment05,
Time_Predict_Payment06, Time_Predict_Payment07, Time_Predict_Payment08, Time_Predict_Payment09, Time_Predict_Payment10, Time_Predict_Payment11),
outputs = c(payment_indicator00_output, payment_mean00_output, payment_indicator01_output, payment_mean01_output,
payment_indicator02_output, payment_mean02_output, payment_indicator03_output, payment_mean03_output,
payment_indicator04_output, payment_mean04_output, payment_indicator05_output, payment_mean05_output,
payment_indicator06_output, payment_mean06_output, payment_indicator07_output, payment_mean07_output,
payment_indicator08_output, payment_mean08_output, payment_indicator09_output, payment_mean09_output,
payment_indicator10_output, payment_mean10_output, payment_indicator11_output, payment_mean11_output))
model %>% compile(optimizer = optimizer_nadam(),
loss = list(payment_indicator00_output = 'binary_crossentropy', payment_mean00_output = 'mse',
payment_indicator01_output = 'binary_crossentropy', payment_mean01_output = 'mse',
payment_indicator02_output = 'binary_crossentropy', payment_mean02_output = 'mse',
payment_indicator03_output = 'binary_crossentropy', payment_mean03_output = 'mse',
payment_indicator04_output = 'binary_crossentropy', payment_mean04_output = 'mse',
payment_indicator05_output = 'binary_crossentropy', payment_mean05_output = 'mse',
payment_indicator06_output = 'binary_crossentropy', payment_mean06_output = 'mse',
payment_indicator07_output = 'binary_crossentropy', payment_mean07_output = 'mse',
payment_indicator08_output = 'binary_crossentropy', payment_mean08_output = 'mse',
payment_indicator09_output = 'binary_crossentropy', payment_mean09_output = 'mse',
payment_indicator10_output = 'binary_crossentropy', payment_mean10_output = 'mse',
payment_indicator11_output = 'binary_crossentropy', payment_mean11_output = 'mse'),
loss_weights = list(payment_indicator00_output = network_weights[1,1], payment_mean00_output = network_weights[1,2],
payment_indicator01_output = network_weights[2,1], payment_mean01_output = network_weights[2,2],
payment_indicator02_output = network_weights[3,1], payment_mean02_output = network_weights[3,2],
payment_indicator03_output = network_weights[4,1], payment_mean03_output = network_weights[4,2],
payment_indicator04_output = network_weights[5,1], payment_mean04_output = network_weights[5,2],
payment_indicator05_output = network_weights[6,1], payment_mean05_output = network_weights[6,2],
payment_indicator06_output = network_weights[7,1], payment_mean06_output = network_weights[7,2],
payment_indicator07_output = network_weights[8,1], payment_mean07_output = network_weights[8,2],
payment_indicator08_output = network_weights[9,1], payment_mean08_output = network_weights[9,2],
payment_indicator09_output = network_weights[10,1], payment_mean09_output = network_weights[10,2],
payment_indicator10_output = network_weights[11,1], payment_mean10_output = network_weights[11,2],
payment_indicator11_output = network_weights[12,1], payment_mean11_output = network_weights[12,2]))
model
}After this code, I ran the following:
library(keras)
library(plyr)
source(file="./NeuralNetworkModels.R")
#########################
### Line of business ###
#########################
LoB <- 1
######################
### Load the data ###
######################
data <- read.table("./Data/data.txt", header=TRUE, sep=";")
data <- data[data$LoB==LoB,]
data$LoB <- as.factor(data$LoB)
data$cc <- as.factor(data$cc)
data$inj_part <- as.factor(data$inj_part)
str(data)
########################
### Data preparation ###
########################
### Features
data$RepDelnew <- pmin(2,data$RepDel) ### cap reporting delay at 2
l0 <- ncol(data)
data$LoBx <- as.integer(as.numeric(data$LoB)-1) ### LoB categorical (we will only use that as artificial embedding for dropout)
cc_transform <- cbind(as.integer(levels(data$cc)),0:(length(levels(data$cc))-1))
data$ccx <- cc_transform[data$cc,2] ### cc categorical
data$AYx <- as.integer(data$AY-1994) ### AY categorical
data$AQx <- as.integer(data$AQ-1) ### AQ categorical
data$agex <- as.integer(levels(cut(data$age, c(14,20,25,30,35,40,45,50,55,60,65,70), right=TRUE, labels=c(0:10))))[cut(data$age, c(14,20,25,30,35,40,45,50,55,60,65,70), right=TRUE, labels=c(0:10))] ### age categorical in 5 years age buckets
inj_part_transform <- cbind(as.integer(levels(data$inj_part)),0:(length(levels(data$inj_part))-1))
data$inj_partx <- inj_part_transform[data$inj_part,2] ### inj_part categorical
data$RepDelx <- as.integer(data$RepDelnew) ### RepDelnew categorical
### Time known at the end of 2005 (time points for which we have the payment information at the end of year 2005)
l1 <- ncol(data)
data[,c("Time_Known00", "Time_Known01", "Time_Known02", "Time_Known03", "Time_Known04", "Time_Known05", "Time_Known06", "Time_Known07", "Time_Known08", "Time_Known09", "Time_Known10", "Time_Known11")] <- 0
for (i in 1:12){
data[data$AY + data$RepDel <= 2006-i,l1+i] <- 1
}
### Past payment info (feature information about past payments, the information at time 11 is never used as a feature and thus neglected)
l2 <- ncol(data)
data[,c("Pay00Info", "Pay01Info", "Pay02Info", "Pay03Info", "Pay04Info", "Pay05Info", "Pay06Info", "Pay07Info", "Pay08Info", "Pay09Info", "Pay10Info")] <- as.integer(0)
for (i in 1:11){
data[,l2+i][data[,l1+i]==1] <- as.integer(levels(cut(data[,8+i][data[,l1+i]==1], c(-Inf,-1,0,5000,20000,100000,Inf), right=TRUE, labels=c(1,0,2,3,4,5))))[cut(data[,8+i][data[,l1+i]==1], c(-Inf,-1,0,5000,20000,100000,Inf), right=TRUE, labels=c(1,0,2,3,4,5))]
}
### Payment indicator (response variable indicating whether we have a payment or not)
l3 <- ncol(data)
data[,c("PayInd00", "PayInd01", "PayInd02", "PayInd03", "PayInd04", "PayInd05", "PayInd06", "PayInd07", "PayInd08", "PayInd09", "PayInd10", "PayInd11")] <- as.integer(0)
for (i in 1:12){
data[,l3+i][data[,l1+i]==1] <- as.integer(data[,8+i][data[,l1+i]==1]>0)
}
### Logarithmic payment (response variable indicating the log size of the payment if there is a payment)
l4 <- ncol(data)
data[,c("LogPay00", "LogPay01", "LogPay02", "LogPay03", "LogPay04", "LogPay05", "LogPay06", "LogPay07", "LogPay08", "LogPay09", "LogPay10", "LogPay11")] <- 0
for (i in 1:12){
data[,l4+i][data[,l1+i]==1 & data[,8+i]>0] <- log(data[,8+i][data[,l1+i]==1 & data[,8+i]>0])
}
### Time points for which we would like to estimate the payment probability (during training these are equal to the time points known)
l5 <- ncol(data)
data[,c("Time_Predict_Indicator00", "Time_Predict_Indicator01", "Time_Predict_Indicator02", "Time_Predict_Indicator03", "Time_Predict_Indicator04", "Time_Predict_Indicator05", "Time_Predict_Indicator06", "Time_Predict_Indicator07", "Time_Predict_Indicator08", "Time_Predict_Indicator09", "Time_Predict_Indicator10", "Time_Predict_Indicator11")] <- 0
l6 <- ncol(data)
data[,(l5+1):l6] <- data[,(l1+1):(l1+12)]
### Time points for which we would like to estimate the payment size (during training these are equal to the time points where we have a payment)
data[,c("Time_Predict_Payment00", "Time_Predict_Payment01", "Time_Predict_Payment02", "Time_Predict_Payment03", "Time_Predict_Payment04", "Time_Predict_Payment05", "Time_Predict_Payment06", "Time_Predict_Payment07", "Time_Predict_Payment08", "Time_Predict_Payment09", "Time_Predict_Payment10", "Time_Predict_Payment11")] <- 0
l7 <- ncol(data)
data[,(l6+1):ncol(data)] <- data[,(l3+1):(l3+12)]
################
### Features ###
################
### Complete data set
features_all <- data.matrix(data[,c((l0+1):l1,(l1+2):l3,(l5+1):l7)]) ### l1+1 is Time_Known00, which is always equal to 1 and therefore omitted
features_all_input <- list(dropout = features_all[,1], cc = features_all[,2], AY = features_all[,3], AQ = features_all[,4], age = features_all[,5], inj_part = features_all[,6], RepDel = features_all[,7],
Time_Known01 = features_all[,8], Time_Known02 = features_all[,9], Time_Known03 = features_all[,10], Time_Known04 = features_all[,11], Time_Known05 = features_all[,12],
Time_Known06 = features_all[,13], Time_Known07 = features_all[,14], Time_Known08 = features_all[,15], Time_Known09 = features_all[,16], Time_Known10 = features_all[,17], Time_Known11 = features_all[,18],
Pay00Info = features_all[,19], Pay01Info = features_all[,20], Pay02Info = features_all[,21], Pay03Info = features_all[,22], Pay04Info = features_all[,23], Pay05Info = features_all[,24],
Pay06Info = features_all[,25], Pay07Info = features_all[,26], Pay08Info = features_all[,27], Pay09Info = features_all[,28], Pay10Info = features_all[,29],
Time_Predict_Indicator00 = features_all[,30], Time_Predict_Indicator01 = features_all[,31], Time_Predict_Indicator02 = features_all[,32], Time_Predict_Indicator03 = features_all[,33], Time_Predict_Indicator04 = features_all[,34], Time_Predict_Indicator05 = features_all[,35],
Time_Predict_Indicator06 = features_all[,36], Time_Predict_Indicator07 = features_all[,37], Time_Predict_Indicator08 = features_all[,38], Time_Predict_Indicator09 = features_all[,39], Time_Predict_Indicator10 = features_all[,40], Time_Predict_Indicator11 = features_all[,41],
Time_Predict_Payment00 = features_all[,42], Time_Predict_Payment01 = features_all[,43], Time_Predict_Payment02 = features_all[,44], Time_Predict_Payment03 = features_all[,45], Time_Predict_Payment04 = features_all[,46], Time_Predict_Payment05 = features_all[,47],
Time_Predict_Payment06 = features_all[,48], Time_Predict_Payment07 = features_all[,49], Time_Predict_Payment08 = features_all[,50], Time_Predict_Payment09 = features_all[,51], Time_Predict_Payment10 = features_all[,52], Time_Predict_Payment11 = features_all[,53])
#################
### Responses ###
#################
### Complete data set
responses_all <- data.matrix(data[,c((l3+1):l5)])
responses_all_input <- list(payment_indicator00_output = responses_all[,1], payment_mean00_output = responses_all[,13],
payment_indicator01_output = responses_all[,2], payment_mean01_output = responses_all[,14],
payment_indicator02_output = responses_all[,3], payment_mean02_output = responses_all[,15],
payment_indicator03_output = responses_all[,4], payment_mean03_output = responses_all[,16],
payment_indicator04_output = responses_all[,5], payment_mean04_output = responses_all[,17],
payment_indicator05_output = responses_all[,6], payment_mean05_output = responses_all[,18],
payment_indicator06_output = responses_all[,7], payment_mean06_output = responses_all[,19],
payment_indicator07_output = responses_all[,8], payment_mean07_output = responses_all[,20],
payment_indicator08_output = responses_all[,9], payment_mean08_output = responses_all[,21],
payment_indicator09_output = responses_all[,10], payment_mean09_output = responses_all[,22],
payment_indicator10_output = responses_all[,11], payment_mean10_output = responses_all[,23],
payment_indicator11_output = responses_all[,12], payment_mean11_output = responses_all[,24])
#######################
### Hyperparameters ###
#######################
### Random seed for Keras
set.seed(100)
seed1 <- sample(1:1000000, 1)
### Starting values of the neural network
starting_values <- array(NA, dim=c(12,2))
for (i in 1:12){
starting_values[i,1] <- mean(data[data$AY+data$RepDel<=2006-i,8+i]>0)
starting_values[i,2] <- mean(log(data[data$AY+data$RepDel<=2006-i,8+i][data[data$AY+data$RepDel<=2006-i,8+i]>0]))
}
### Dropout rates
dropout_rates <- 1/c(2:11)
### Weights
network_weights <- array(NA,dim=c(12,2))
for (i in 1:12){
a <- as.numeric(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0)
b <- nrow(data)-length(a)
network_weights[i,1] <- -mean(c(a*log(mean(a))+(1-a)*log(1-mean(a)),rep(0,b)))
a <- log(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i][data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0])
b <- nrow(data)-length(a)
network_weights[i,2] <- mean(c((a-mean(a))^2,rep(0,b)))
}
network_weights <- 1/network_weights
### Numbers of neurons
neurons <- c(40,30,10)
### Epochs
epochs <- as.numeric(read.table(paste("./Results/NumbersOfEpochs/FirstTrainingStep/LoB",LoB,"/Number_of_Epochs_Embedding.txt", sep=""), header=TRUE, sep=";",row.names=1))
### Batchsize
batchsize <- 10000
###############################################
### Application of the neural network model ###
###############################################
k_clear_session()
model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)
fit = model %>% fit(x = features_all_input, y = responses_all_input, epochs = epochs, batch_size = batchsize, verbose = 0)
#########################
### Store the results ###
#########################
### Determine which of the model weights are embeddings
a <- get_weights(model)
embeddings <- c()
for (i in 1:length(a)){
if (nrow(a[[i]])>1 & is.element(nrow(a[[i]]),15:26)==FALSE & is.na(ncol(a[[i]]))==FALSE & ncol(a[[i]])==1){
embeddings <- c(embeddings,i)
}
}
write.table(embeddings, file=paste("./Results/Embeddings/LoB",LoB,"/Embedding_Weights_Indicator.txt", sep=""), sep=";", row.names=TRUE, col.names=NA)
### Store the embeddings
for (i in embeddings){
write.table(a[[i]], file=paste("./Results/Embeddings/LoB",LoB,"/Embedding_Weights_",i,".txt", sep=""), sep=";", row.names=TRUE, col.names=NA)
}I started to receive the error from the model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights) line and below.
t-kalinowski commented
Can you please condense or isolate the issue more?
Some helpful guidelines are described here: https://www.tidyverse.org/help/ In particulate, please note the second bullet point, starting with "Second, you need to make it minimal..."
Another take that gives a few more concrete recommendations on what an MRE is:
https://robjhyndman.com/hyndsight/minimal-reproducible-examples/#creating-an-mre
Waselia commented
@t-kalinowski Sure! Sorry for my misunderstanding, I'm new to the coding and machine learning. Here's a more condensed code, I only put the first line of code that seems to have the error ValueError: Unrecognized keyword arguments passed to Embedding: {'input_length': 1, 'weights': [array([[1.]])]}. From what I know, it is because the Embedding in Python only has 2 arguments input_dim and output_dim, whilst R allows to have input_length and weight.
Please import the dataset as well: data.txt
library(keras)
library(plyr)
library(reticulate)
library(tensorflow)
neural_network_train_embedding <- function(seed1, neurons, dropout_rates, starting_values, network_weights){
### Seed
# use_session_with_seed(seed1)
set_random_seed(seed1)
### Dropout
dropout_1_1 <- layer_dropout(rate = dropout_rates[1])
### Inputs
dropout <- layer_input(shape = c(1), dtype = 'int32', name = 'dropout')
### Embeddings
dropout_embed <- dropout %>%
layer_embedding(input_dim = 1, output_dim = 1, trainable=FALSE, input_length = 1,
weights=list(matrix(c(1)))) %>%
layer_flatten
### Keras model
model <- keras_model(inputs = c(dropout))
model
}
#######################
### Hyperparameters ###
#######################
### Random seed for Keras
set.seed(100)
seed1 <- sample(1:1000000, 1)
### Starting values of the neural network
starting_values <- array(NA, dim=c(12,2))
for (i in 1:12){
starting_values[i,1] <- mean(data[data$AY+data$RepDel<=2006-i,8+i]>0)
starting_values[i,2] <- mean(log(data[data$AY+data$RepDel<=2006-i,8+i][data[data$AY+data$RepDel<=2006-i,8+i]>0]))
}
### Dropout rates
dropout_rates <- 1/c(2:11)
### Weights
network_weights <- array(NA,dim=c(12,2))
for (i in 1:12){
a <- as.numeric(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0)
b <- nrow(data)-length(a)
network_weights[i,1] <- -mean(c(a*log(mean(a))+(1-a)*log(1-mean(a)),rep(0,b)))
a <- log(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i][data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0])
b <- nrow(data)-length(a)
network_weights[i,2] <- mean(c((a-mean(a))^2,rep(0,b)))
}
network_weights <- 1/network_weights
### Numbers of neurons
neurons <- c(40,30,10)
### Epochs
epochs <- 30
### Batchsize
batchsize <- 10000
###############################################
### Application of the neural network model ###
###############################################
k_clear_session()
model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)And I received the error from here
t-kalinowski commented
How did you read in 'data'? Would it be possible to dump it out using dput() and include it in the MRE?
Waselia commented
@t-kalinowski I use the read.table() :
data <- read.table("D:/data.txt", quote="\"", comment.char="")
View(data)Or using `dput()`:
structure(list(ClNr = c(1L, 5L, 7L, 14L, 22L, 29L, 36L, 40L,
41L, 44L, 54L, 57L, 61L, 64L, 71L, 75L, 77L, 81L, 85L, 87L, 93L,
98L, 104L, 110L, 112L, 121L, 125L, 130L, 138L, 155L, 158L, 162L,
166L, 168L, 173L, 177L, 186L, 190L, 193L, 194L, 195L, 196L, 199L,
213L), LoB = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L
), cc = c(18L, 19L, 6L, 6L, 46L, 27L, 19L, 45L, 17L, 29L, 19L,
17L, 22L, 19L, 46L, 17L, 47L, 17L, 17L, 50L, 41L, 6L, 20L, 53L,
20L, 17L, 14L, 14L, 48L, 47L, 53L, 22L, 19L, 37L, 20L, 45L, 47L,
17L, 14L, 17L, 45L, 6L, 13L, 6L), AY = c(1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L), AQ = c(3L, 1L, 3L, 1L, 3L,
1L, 1L, 1L, 3L, 3L, 3L, 4L, 1L, 3L, 4L, 2L, 1L, 1L, 3L, 4L, 4L,
3L, 4L, 1L, 4L, 3L, 3L, 1L, 1L, 4L, 4L, 2L, 1L, 4L, 1L, 3L, 2L,
1L, 2L, 4L, 2L, 2L, 3L, 1L), age = c(40L, 35L, 50L, 69L, 24L,
23L, 39L, 24L, 38L, 51L, 17L, 36L, 23L, 26L, 44L, 34L, 47L, 17L,
40L, 42L, 26L, 52L, 46L, 51L, 49L, 43L, 19L, 48L, 32L, 35L, 37L,
40L, 43L, 42L, 17L, 35L, 19L, 46L, 46L, 41L, 19L, 21L, 22L, 29L
), inj_part = c(52L, 61L, 53L, 23L, 35L, 60L, 36L, 15L, 41L,
53L, 70L, 36L, 53L, 12L, 53L, 35L, 53L, 55L, 54L, 36L, 12L, 36L,
41L, 42L, 37L, 34L, 12L, 12L, 11L, 36L, 50L, 53L, 55L, 35L, 51L,
64L, 34L, 50L, 36L, 53L, 12L, 36L, 71L, 36L), RepDel = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L), Pay00 = c(1341L,
381L, 160L, 694L, 457L, 2005L, 554L, 400L, 279L, 731L, 215L,
515L, 255L, 82L, 0L, 574L, 1237L, 37L, 616L, 0L, 61L, 636L, 790L,
3370L, 460L, 361L, 153L, 88L, 85L, 0L, 61L, 429L, 24L, 313L,
172L, 868L, 8757L, 569L, 92L, 725L, 145L, 262L, 421L, 318L),
Pay01 = c(1028L, 0L, 0L, 0L, 0L, 959L, 0L, 0L, 0L, 0L, 0L,
344L, 0L, 0L, 577L, 0L, 0L, 0L, 0L, 146L, 0L, 0L, 913L, 0L,
0L, 0L, 0L, 0L, 0L, 145L, 0L, 0L, 0L, 0L, 0L, 0L, 6312L,
0L, 0L, 430L, 0L, 0L, 0L, 0L), Pay02 = c(548L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 563L, 0L,
0L, 0L, 0L, 0L, 426L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 5520L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay03 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 3147L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay04 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay05 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay06 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay07 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay08 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay09 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay10 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay11 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), RY = c(1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1995L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1995L, 1994L, 1994L, 1994L,
1994L, 1994L, 1994L, 1994L, 1994L, 1995L, 1994L, 1994L, 1994L,
1994L, 1994L, 1995L, 1994L, 1994L, 1994L, 1994L), RepDelnew = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L), LoBx = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), ccx = c(16L,
17L, 4L, 4L, 43L, 25L, 17L, 42L, 15L, 27L, 17L, 15L, 20L,
17L, 43L, 15L, 44L, 15L, 15L, 47L, 38L, 4L, 18L, 50L, 18L,
15L, 12L, 12L, 45L, 44L, 50L, 20L, 17L, 34L, 18L, 42L, 44L,
15L, 12L, 15L, 42L, 4L, 11L, 4L), AYx = c(0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), AQx = c(2L, 0L,
2L, 0L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 3L, 0L, 2L, 3L, 1L, 0L,
0L, 2L, 3L, 3L, 2L, 3L, 0L, 3L, 2L, 2L, 0L, 0L, 3L, 3L, 1L,
0L, 3L, 0L, 2L, 1L, 0L, 1L, 3L, 1L, 1L, 2L, 0L), agex = c(4L,
3L, 6L, 10L, 1L, 1L, 4L, 1L, 4L, 7L, 0L, 4L, 1L, 2L, 5L,
3L, 6L, 0L, 4L, 5L, 2L, 7L, 6L, 7L, 6L, 5L, 0L, 6L, 3L, 3L,
4L, 4L, 5L, 5L, 0L, 3L, 0L, 6L, 6L, 5L, 0L, 1L, 1L, 2L),
inj_partx = c(30L, 36L, 31L, 10L, 19L, 35L, 20L, 5L, 23L,
31L, 41L, 20L, 31L, 2L, 31L, 19L, 31L, 33L, 32L, 20L, 2L,
20L, 23L, 24L, 21L, 18L, 2L, 2L, 1L, 20L, 28L, 31L, 33L,
19L, 29L, 39L, 18L, 28L, 20L, 31L, 2L, 20L, 42L, 20L), RepDelx = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L), Time_Known00 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known01 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known02 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known03 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known04 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known05 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known06 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known07 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known08 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known09 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known10 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Known11 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L), Pay00Info = c(2L,
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L,
2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 3L, 2L, 2L, 2L, 2L, 2L, 2L, 2L), Pay01Info = c(2L,
0L, 0L, 0L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 0L, 0L, 2L, 0L,
0L, 0L, 0L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 0L,
0L, 0L, 0L, 0L, 0L, 3L, 0L, 0L, 2L, 0L, 0L, 0L, 0L), Pay02Info = c(2L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
2L, 0L, 0L, 0L, 0L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 3L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay03Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay04Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay05Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay06Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay07Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay08Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay09Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Pay10Info = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd00 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L,
1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), PayInd01 = c(1L,
0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L,
0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L, 0L, 0L, 0L), PayInd02 = c(1L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd03 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd04 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd05 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd06 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd07 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd08 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd09 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd10 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), PayInd11 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay00 = c(7.20117088328168,
5.9427993751267, 5.07517381523383, 6.5424719605068, 6.12468339089421,
7.60339933974067, 6.31716468674728, 5.99146454710798, 5.63121178182137,
6.59441345974978, 5.37063802812766, 6.24416690066374, 5.54126354515843,
4.40671924726425, 0, 6.35262939631957, 7.12044437239249,
3.61091791264422, 6.42324696353352, 0, 4.11087386417331,
6.45519856334012, 6.67203294546107, 8.12266802334641, 6.13122648948314,
5.88887795833288, 5.03043792139244, 4.47733681447821, 4.44265125649032,
0, 4.11087386417331, 6.06145691892802, 3.17805383034795,
5.74620319054015, 5.14749447681345, 6.76619171466035, 9.07760865952222,
6.34388043412633, 4.52178857704904, 6.58617165485467, 4.97673374242057,
5.5683445037611, 6.04263283368238, 5.76205138278018), LogPay01 = c(6.93537044601511,
0, 0, 0, 0, 6.86589107488344, 0, 0, 0, 0, 0, 5.8406416573734,
0, 0, 6.3578422665081, 0, 0, 0, 0, 4.98360662170834, 0, 0,
6.81673588059497, 0, 0, 0, 0, 0, 0, 4.97673374242057, 0,
0, 0, 0, 0, 0, 8.75020786252571, 0, 0, 6.06378520868761,
0, 0, 0, 0), LogPay02 = c(6.30627528694802, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6.33327962813969, 0, 0, 0,
0, 0, 6.05443934626937, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 8.61613313927114, 0, 0, 0, 0, 0, 0, 0), LogPay03 = c(0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8.05420489706441,
0, 0, 0, 0, 0, 0, 0), LogPay04 = c(0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay05 = c(0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay06 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay07 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay08 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay09 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay10 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), LogPay11 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Indicator00 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator01 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator02 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator03 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator04 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator05 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator06 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator07 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator08 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator09 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator10 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Indicator11 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 1L,
1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L), Time_Predict_Payment00 = c(1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L,
1L, 1L, 1L, 0L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 0L, 1L,
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), Time_Predict_Payment01 = c(1L,
0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L,
0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 1L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 1L, 0L, 0L, 0L, 0L), Time_Predict_Payment02 = c(1L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
1L, 0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment03 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment04 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment05 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment06 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment07 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment08 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment09 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment10 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L), Time_Predict_Payment11 = c(0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L,
0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L)), class = "data.frame", row.names = c("1",
"3", "4", "6", "9", "11", "15", "17", "18", "20", "23", "26",
"29", "30", "32", "34", "36", "38", "39", "41", "43", "46", "51",
"54", "55", "58", "61", "63", "65", "70", "73", "76", "80", "81",
"83", "84", "88", "89", "91", "92", "93", "94", "97", "100"))t-kalinowski commented
The input_length argument was removed from tf.keras.layers.Embedding() with the transition to Keras 3. In order to run this code without updating it, you need to use legacy keras. You can still use the legacy keras with the latest TensorFlow release version (2.16) by installing the tf_keras pip package and setting the env var TF_USE_LEGACY_KERAS=1 before loading tensorflow
I recommend creating a project specific venv for this legacy project:
virtualenv_create("./venv", version = ">=3.9,<=3.11", packages = c(
"tensorflow", "tf_keras" # legacy keras module
))(Reticulate will automatically discover a "./venv" in the current working directory, see here for how reticulate selects a Python installation)
Then, before initializing TensorFlow/Keras, make sure the env var is set
Sys.setenv(TF_USE_LEGACY_KERAS=1)
library(tensorflow)
library(keras)
# use_virtualenv(<path/to/venv>) # if you don't use the default auto-discoverable "./venv" location
# <rest of code>Waselia commented
Hi, thank you so much for helping me out. However, I tried to run the code and received this error:
virtualenv_create("./venv", version = ">=3.9,<=3.11", packages = c("tensorflow", "tf_keras")) # legacy keras module
Error in stop_no_virtualenv_starter(version = version, python = python) :
Suitable Python installation for creating a venv not found.
Requested version constraint: >=3.9,<=3.11
Please install Python with one of following methods:
- https://www.python.org/downloads/
- reticulate::install_python(version = '<version>')Here's a full version of the code:
library(reticulate)
virtualenv_create("./venv", version = ">=3.9,<=3.11", packages = c("tensorflow", "tf_keras")) # legacy keras module
Sys.setenv(TF_USE_LEGACY_KERAS=1)
library(tensorflow)
library(keras)
# use_virtualenv(<path/to/venv>) # if you don't use the default auto-discoverable "./venv" location
# <rest of code>I hope I got it right. And my current Python version is 3.11.9.
t-kalinowski commented
Suitable Python installation for creating a venv not found.
Reticulate is not finding a Python that it can use to create the venv with. You can remedy by installing Python in one of the common locations by doing one of the options suggested in the error message (e.g., calling reticulate::install_python()), or by explicitly providing an absolute path to a Python binary in virtualenv_create(), like this:
virtualenv_create("./venv", python = "/path/to/python.exe", packages = ...)Waselia commented
I have tried reticulate::install_python(), which installs python 3.11.10 for me, but I still receive the same error for the model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights) code:
> model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)
2024-04-08 13:12:28.097193: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
ValueError: Unrecognized keyword arguments passed to Embedding: {'input_length': 1, 'weights': [array([[1.]])]}
Run `reticulate::py_last_error()` for details.t-kalinowski commented
MRE:
library(keras)
layer_input(1, dtype = "int32") |>
layer_embedding(
input_dim = 1,
output_dim = 1,
input_length = 1,
weights = list(matrix(1))
)
#> Unrecognized keyword arguments passed to Embedding: {'input_length': 1,
#> 'weights': [array([[1.]])]}Created on 2024-04-08 with reprex v2.1.0
Fix:
# reticulate::py_install("tf_keras")
Sys.setenv(TF_USE_LEGACY_KERAS=1)
library(keras)
layer_input(1, dtype = "int32") |>
layer_embedding(
input_dim = 1,
output_dim = 1,
input_length = 1,
weights = list(matrix(1))
)Created on 2024-04-08 with reprex v2.1.0
If you're having installation issues, can you please open a new issue with the information requested in https://github.com/rstudio/keras/blob/main/.github/ISSUE_TEMPLATE/installation-issue.md (i.e., the exact commands you are running, and the output from install_tensorflow(), py_config(), sessionInfo())
Waselia commented
Hi, thank you so much for your help. The problem with the layer_embedding code has been solved. I now received a new error with thelayer_multiply code, can you please help me out? From what I know, the inputs can be missing as shown in here.
library(keras)
library(plyr)
library(reticulate)
library(tensorflow)
neural_network_train_embedding <- function(seed1, neurons, dropout_rates, starting_values, network_weights){
### Seed
# use_session_with_seed(seed1)
set_random_seed(seed1)
payment_indicator00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator00_pre2 <- list(payment_indicator00_pre, features_output) %>%
layer_concatenate(name = "pay_ind00") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(log(starting_values[1,1]/(1-starting_values[1,1])),dim=c(1))))
payment_indicator00 <- list(payment_indicator00_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
Time_Predict_Indicator00 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator00')
payment_indicator00_output <- list(payment_indicator00, Time_Predict_Indicator00) %>%
layer_multiply(name = 'payment_indicator00_output')
### Keras model
model <- keras_model(inputs = c(dropout))
model
}
#######################
### Hyperparameters ###
#######################
### Random seed for Keras
set.seed(100)
seed1 <- sample(1:1000000, 1)
### Starting values of the neural network
starting_values <- array(NA, dim=c(12,2))
for (i in 1:12){
starting_values[i,1] <- mean(data[data$AY+data$RepDel<=2006-i,8+i]>0)
starting_values[i,2] <- mean(log(data[data$AY+data$RepDel<=2006-i,8+i][data[data$AY+data$RepDel<=2006-i,8+i]>0]))
}
### Dropout rates
dropout_rates <- 1/c(2:11)
### Weights
network_weights <- array(NA,dim=c(12,2))
for (i in 1:12){
a <- as.numeric(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0)
b <- nrow(data)-length(a)
network_weights[i,1] <- -mean(c(a*log(mean(a))+(1-a)*log(1-mean(a)),rep(0,b)))
a <- log(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i][data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0])
b <- nrow(data)-length(a)
network_weights[i,2] <- mean(c((a-mean(a))^2,rep(0,b)))
}
network_weights <- 1/network_weights
### Numbers of neurons
neurons <- c(40,30,10)
### Epochs
epochs <- 30
### Batchsize
batchsize <- 10000
###############################################
### Application of the neural network model ###
###############################################
k_clear_session()
model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)I received the following error:
model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)
Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
KeyError: 0
Run `reticulate::py_last_error()` for details.
6. stop(<environment>)
5. (structure(function (inputs, ...)
{
cl <- sys.call()
cl[[1L]] <- list2 ...
4. do.call(keras$layers$multiply, c(list(inputs), dots$named))
3. layer_multiply(., name = "payment_indicator00_output")
2. list(payment_indicator00, Time_Predict_Indicator00) %>% layer_multiply(name = "payment_indicator00_output")
1. neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)Thanks again for your support!
Waselia commented
Hi, thanks for your help. The problem with the layer_embedding has been solved. I now have another issue with the layer_multiply code:
library(keras)
library(plyr)
library(reticulate)
library(tensorflow)
neural_network_train_embedding <- function(seed1, neurons, dropout_rates, starting_values, network_weights){
### Seed
# use_session_with_seed(seed1)
set_random_seed(seed1)
hidden_0_pre <- features_input %>%
layer_dense(units=neurons[1], activation='linear')
hidden_0 <- list(hidden_0_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator00_pre2 <- list(payment_indicator00_pre, features_output) %>%
layer_concatenate(name = "pay_ind00") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(log(starting_values[1,1]/(1-starting_values[1,1])),dim=c(1))))
payment_indicator00 <- list(payment_indicator00_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
Time_Predict_Indicator00 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator00')
payment_indicator00_output <- list(payment_indicator00, Time_Predict_Indicator00) %>%
layer_multiply(name = 'payment_indicator00_output')
### Keras model
model <- keras_model(inputs = c(dropout))
model
}
#######################
### Hyperparameters ###
#######################
### Random seed for Keras
set.seed(100)
seed1 <- sample(1:1000000, 1)
### Starting values of the neural network
starting_values <- array(NA, dim=c(12,2))
for (i in 1:12){
starting_values[i,1] <- mean(data[data$AY+data$RepDel<=2006-i,8+i]>0)
starting_values[i,2] <- mean(log(data[data$AY+data$RepDel<=2006-i,8+i][data[data$AY+data$RepDel<=2006-i,8+i]>0]))
}
### Dropout rates
dropout_rates <- 1/c(2:11)
### Weights
network_weights <- array(NA,dim=c(12,2))
for (i in 1:12){
a <- as.numeric(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0)
b <- nrow(data)-length(a)
network_weights[i,1] <- -mean(c(a*log(mean(a))+(1-a)*log(1-mean(a)),rep(0,b)))
a <- log(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i][data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0])
b <- nrow(data)-length(a)
network_weights[i,2] <- mean(c((a-mean(a))^2,rep(0,b)))
}
network_weights <- 1/network_weights
### Numbers of neurons
neurons <- c(40,30,10)
### Epochs
epochs <- 30
### Batchsize
batchsize <- 10000
###############################################
### Application of the neural network model ###
###############################################
k_clear_session()
model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)I received this error:
> model <- neural_network_train_embedding(seed1, neurons, dropout_rates, starting_values, network_weights)
Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
KeyError: 0
Run `reticulate::py_last_error()` for details.
6. stop(<environment>)
5. (structure(function (inputs, ...)
{
cl <- sys.call()
cl[[1L]] <- list2 ...
4. do.call(keras$layers$multiply, c(list(inputs), dots$named))
3. layer_multiply(., name = "payment_indicator00_output")
2. list(payment_indicator00, Time_Predict_Indicator00) %>% layer_multiply(name = "payment_indicator00_output")
1. neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)From what I know, The inputs in the layer_multiply can be missing, as shown in here.
Thanks again for your support!
Waselia commented
Hi, thanks for your help. The problem with the `layer_embedding` has been
solved. I now have another issue with the `layer_multiply` code, which
gives me the `KeyError: 0` error:
```r
library(keras)
library(plyr)
library(reticulate)
library(tensorflow)
neural_network_train_embedding <- function(seed1, neurons, dropout_rates,
starting_values, network_weights){
### Seed
# use_session_with_seed(seed1)
set_random_seed(seed1)
hidden_0_pre <- features_input %>%
layer_dense(units=neurons[1], activation='linear')
hidden_0 <- list(hidden_0_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights =
list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator00_pre2 <- list(payment_indicator00_pre,
features_output) %>%
layer_concatenate(name = "pay_ind00") %>%
layer_dense(units = 1, activation='linear',
weights =
list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(log(starting_values[1,1]/(1-starting_values[1,1])),dim=c(1))))
payment_indicator00 <- list(payment_indicator00_pre2,
AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
Time_Predict_Indicator00 <- layer_input(shape = c(1), dtype = 'float32',
name = 'Time_Predict_Indicator00')
payment_indicator00_output <- list(payment_indicator00,
Time_Predict_Indicator00) %>%
layer_multiply(name = 'payment_indicator00_output')
### Keras model
model <- keras_model(inputs = c(dropout))
model
}
#######################
### Hyperparameters ###
#######################
### Random seed for Keras
set.seed(100)
seed1 <- sample(1:1000000, 1)
### Starting values of the neural network
starting_values <- array(NA, dim=c(12,2))
for (i in 1:12){
starting_values[i,1] <- mean(data[data$AY+data$RepDel<=2006-i,8+i]>0)
starting_values[i,2] <-
mean(log(data[data$AY+data$RepDel<=2006-i,8+i][data[data$AY+data$RepDel<=2006-i,8+i]>0]))
}
### Dropout rates
dropout_rates <- 1/c(2:11)
### Weights
network_weights <- array(NA,dim=c(12,2))
for (i in 1:12){
a <- as.numeric(data[data$AY+data$RepDel<=2006-i & data[,29+i]==1,8+i]>0)
b <- nrow(data)-length(a)
network_weights[i,1] <-
-mean(c(a*log(mean(a))+(1-a)*log(1-mean(a)),rep(0,b)))
a <- log(data[data$AY+data$RepDel<=2006-i &
data[,29+i]==1,8+i][data[data$AY+data$RepDel<=2006-i &
data[,29+i]==1,8+i]>0])
b <- nrow(data)-length(a)
network_weights[i,2] <- mean(c((a-mean(a))^2,rep(0,b)))
}
network_weights <- 1/network_weights
### Numbers of neurons
neurons <- c(40,30,10)
### Epochs
epochs <- 30
### Batchsize
batchsize <- 10000
###############################################
### Application of the neural network model ###
###############################################
k_clear_session()
model <- neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)
```
I received this error:
```r
model <- neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)
Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
KeyError: 0
Run `reticulate::py_last_error()` for details.
6. stop(<environment>)
5. (structure(function (inputs, ...)
{
cl <- sys.call()
cl[[1L]] <- list2 ...
4. do.call(keras$layers$multiply, c(list(inputs), dots$named))
3. layer_multiply(., name = "payment_indicator00_output")
2. list(payment_indicator00, Time_Predict_Indicator00) %>%
layer_multiply(name = "payment_indicator00_output")
1. neural_network_train_embedding(seed1, neurons, dropout_rates,
starting_values, network_weights)
```
From what I know, The `inputs` in the `layer_multiply` can be missing, as
shown in [here](
https://tensorflow.rstudio.com/reference/keras/layer_multiply).
Thanks again for your support!
Vào Th 2, 8 thg 4, 2024 vào lúc 19:33 Tomasz Kalinowski <
***@***.***> đã viết:
… MRE:
library(keras)
layer_input(1, dtype = "int32") |>
layer_embedding(
input_dim = 1,
output_dim = 1,
input_length = 1,
weights = list(matrix(1))
)#> Unrecognized keyword arguments passed to Embedding: {'input_length': 1,#> 'weights': [array([[1.]])]}
Created on 2024-04-08 with reprex v2.1.0 <https://reprex.tidyverse.org>
Fix:
# reticulate::py_install("tf_keras")
Sys.setenv(TF_USE_LEGACY_KERAS=1)
library(keras)
layer_input(1, dtype = "int32") |>
layer_embedding(
input_dim = 1,
output_dim = 1,
input_length = 1,
weights = list(array(1, c(1, 1)))
)
Created on 2024-04-08 with reprex v2.1.0 <https://reprex.tidyverse.org>
If you're having installation issues, can you please open a new issue, can
you please open a new issue with the information requested in
https://github.com/rstudio/keras/blob/main/.github/ISSUE_TEMPLATE/installation-issue.md
(i.e., the exact commands you are running, and the output from install_tensorflow(),
py_config(), sessionInfo()`)
—
Reply to this email directly, view it on GitHub
<#1427 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/BHUK4HZF5SESACKE5SYOATDY4KFDLAVCNFSM6AAAAABF2TLZG2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANBSGYZTAMRVHA>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
t-kalinowski commented
Can you please post an MRE (the above is not minimal enough).
You can call debug(layer_multiply), then observe the arguments passed into it, and attempt to recreate the error with just a similar bare call at the top level.
Waselia commented
Hi, sorry for the late response. That is the shortest MRE I could provide. I received error for the following line of code: payment_indicator00_output <- list(payment_indicator00, Time_Predict_Indicator00) %>% layer_multiply(name = 'payment_indicator00_output')
I also tried payment_indicator00_output <- layer_multiply(list(payment_indicator00, Time_Predict_Indicator00), name = 'payment_indicator00_output')
But I still received the error like I posted in the previous post.
t-kalinowski commented
Can you please make your MRE self contained? Please use the reprex package: https://reprex.tidyverse.org
Waselia commented
Hi, here's the MRE. Thanks again for your help!
library(keras)
library(plyr)
hidden_0_pre <- features_input %>%
layer_dense(units=neurons[1], activation='linear')
hidden_0 <- list(hidden_0_pre, AY_embed_input_NN) %>%
layer_add %>%
layer_dense(units=neurons[1], activation='tanh', trainable = FALSE,
weights = list(diag(1,neurons[1]),array(0,dim=c(neurons[1])))) %>%
layer_dense(units=neurons[2], activation='tanh')
payment_indicator00_pre <- hidden_0 %>%
layer_dense(units=neurons[3], activation='tanh')
payment_indicator00_pre2 <- list(payment_indicator00_pre, features_output) %>%
layer_concatenate(name = "pay_ind00") %>%
layer_dense(units = 1, activation='linear',
weights = list(array(rep(0,neurons[3]+5),dim=c(neurons[3]+5,1)),array(log(starting_values[1,1]/(1-starting_values[1,1])),dim=c(1))))
payment_indicator00 <- list(payment_indicator00_pre2, AY_output_indicator) %>%
layer_add %>%
layer_dense(units = 1, activation = 'sigmoid', trainable = FALSE,
weights = list(array(1,dim=c(1,1)),array(0,dim=c(1))))
Time_Predict_Indicator00 <- layer_input(shape = c(1), dtype = 'float32', name = 'Time_Predict_Indicator00')
payment_indicator00_output <- list(payment_indicator00, Time_Predict_Indicator00) %>%
layer_multiply(name = 'payment_indicator00_output')t-kalinowski commented
When I try to run the reprex, I see:
Error:
! object 'neurons' not found
Run `rlang::last_trace()` to see where the error occurred.
> rlang::last_trace()
<error/rlang_error>
Error:
! object 'neurons' not found
---
Backtrace:
▆
1. ├─base::source("~/.active-rstudio-document")
2. │ ├─base::withVisible(eval(ei, envir))
3. │ └─base::eval(ei, envir)
4. │ └─base::eval(ei, envir)
5. ├─features_input %>% ... at ~/.active-rstudio-document:4:1
6. └─keras::layer_dense(., units = neurons[1], activation = "linear")
7. └─keras::create_layer(...)Please use reprex::reprex() to generate the MRE
github-actions commented
Automatically closed because there has not been a response for 30 days. When you're ready to work on this further, please comment here and the issue will automatically reopen.