- Music data visualization
- Shiny Weather Forecast
- Basic neural network implementation using Rcpp and RcppArmadillp
- Restricted Boltzmann Machines in R
- Scraping IMDB data and basic recommendation algorithm
Shiny App for music data visualization using REchoNest and twitteR
- to make ui user friendly.(current shiny ui is very messy.)
The above Shiny App is to plot weather forecast data of India. To get weather forecast data, I have written a R package ROpenWeatherMap. ROpenWeatherMap is a R wrapper to openweathermap API through which we can access 5 day weather forecast data at any location or city.It includes weather data every 3 hours. I have used googleGvis as a plotting library. Before running this Shiny App you will have to install ROpenWeatherMap package (Installation instructions). To get OpenWeatherMAP API key, sign up here
To run this app:
api_key="YOUR API KEY"
library(shiny)
runApp("/Path to Shiny weather forecast folder")
The 'Rcpp' package provides R functions as well as C++ classes which offer a seamless integration of R and C++.RcppArmadillo provides an interface from R to and from Armadillo by utilising the Rcpp R/C++ interface library.
library(Rcpp)
library(RcppArmadillo)
source(neural_network_2.cpp)
### to call constructor of neural_network class
### new(neural_network,num_input_neurons,num_hidden_neurons,num_output_neurons)
nn=new(neural_network,10,64,1)
### to set learning parameters
### nn$set_learning_parameters(learning_rate,momentum)
nn$set_learning_parameters(0.01,0.01)
### to set maximum number of epochs
### nn$set_max_epochs(max_epochs)
nn$set_max_epochs(2000)To train neural network
### to train neural network, first convert input data into list as follows
### train=matrix(as.numeric(unlist(data2)),nrow=nrow(data2))
### train=list(data=training_part,label=matrix(labels))
### sample code is given in sample.R file
### nn$train_network(train_data_list,gen_data_list,val_data_list)
nn$train_network(train,gen,val)To save or load weights
### to save weights
nn$save_weights()
### to load weights
nn$load_weights()To predict on test datasets
### first convert test data into matrix
### test=matrix(as.numeric(unlist(test_data)),nrow=nrow(test_data))
### to predict on test data matrix
pred=nn$predict(test)- to optimize C++ code (by removing for loops and using RcppArmadillo's builtin functions for matrix multiplication)
Sample codes are given in sample.R and sample2.R
### to load RBM code
library(Rcpp)
library(RcppArmadillo)
source("rbm.cpp")
### to initialize RBM object
new(rbm,num_visible,num_hidden,learning_rate,max_epochs)
### to train RBM
d$train(train_matrix)
### to predict
d1=d$run_visible(test_matrix)
d$run_hidden(d1)
### to save weights
d$save_weights()
### to load weights
d$load_weights()- add daydream function
Blog Link: Scraping IMDB data and basic recommendation algorithm