patrick-kidger/torchcde

Issues

• Please consider create conda distribution (maybe conda-forge?)

  #49 opened a year ago by Saltsmart
  1

• Overfitting

  #53 opened a year ago by ayushpatnaikgit
  0

• Seq2seq forecasting: adding temporal information and extrapolation

  #52 opened a year ago by ClaudiaShu
  0

• Masking Coefficients?

  #51 opened a year ago by AshCher51
  2

• Prediction of irregular time series

  #50 opened a year ago by NonoMalpi
  1

• About how to use CDE in Variational Autoencoders?

  #43 opened a year ago by Saltsmart
  1

• Integrate the ODE function of the CDE system to infinity

  #48 opened 2 years ago by Basicallyimwell
  2

• much slower when using torchsde as backend.

  #47 opened 2 years ago by Fei-u
  2

• Sequence outputs from Neural ODE (similar to 'many to many' RNN)?

  #40 opened 3 years ago by dactylogram
  3

• Very slow training with market data, normal?

  #46 opened 3 years ago by Karlheinzniebuhr
  1

• piping in & predicting arbitrary streams of values

  #45 opened 3 years ago by boughtbot
  0

• Integration to pytorch lighting pipeline

  #44 opened 3 years ago by Karlheinzniebuhr
  1

• About how to use on seq2seq works

  #39 opened 3 years ago by caisikai
  5

• Comparison to alternative ODE models

  #42 opened 3 years ago by denizetkar
  4

• CDEs with Image Data

  #41 opened 3 years ago by AshCher51
  1

• Consider opening a GH discussions

  #37 opened 3 years ago by darleybarreto
  13

• Online prediction tasks needs examples

  #38 opened 3 years ago by denizetkar
  4

• publish `torchcde` on PyPI

  #29 opened 3 years ago by Zymrael
  1

• Breaking install

  #27 opened 3 years ago by howardjp
  4

• Add log-ODE example.

  #18 opened 3 years ago by patrick-kidger
  0

• setup.py and torchsde

  #26 opened 4 years ago by oizin
  1

• How to use log_ode? What does "window" represent time or number of points?

  #25 opened 4 years ago by Krasner
  3

• generation example

  #23 opened 4 years ago by xiaoiker
  2

• Replace the fill-backwards, given missing data in the interpolations, with something better

  #22 opened 4 years ago by patrick-kidger
  1

• TupleControl + computed parameters

  #21 opened 4 years ago by patrick-kidger
  1

• Fix interpolation documentation

  #17 opened 4 years ago by patrick-kidger
  1

• Fix logsignatures to divide by window length

  #12 opened 4 years ago by patrick-kidger
  7

• Add Signatory as an optional dependency

  #11 opened 4 years ago by patrick-kidger
  1

• Time channel unaltered after interpolation and automatic padding when missing values

  #20 opened 4 years ago by JoaquinGajardo
  6

• Allow multiple channel-like dimensions?

  #16 opened 4 years ago by patrick-kidger
  0

• Allow tupled z0

  #2 opened 4 years ago by patrick-kidger
  0

• Natural cubic splines shouldn't fill in just the final value of a sequence, but all missing values at the end. (Same at the start.)

  #14 opened 4 years ago by patrick-kidger
  0

• Basic prediction problem.

  #9 opened 4 years ago by andrewcztrack
  2

• Allow cubic splines with a kink + zero derivative at the end.

  #13 opened 4 years ago by patrick-kidger
  0

• Update README with references to causality-and-controls papers in the appropriate places.

  #5 opened 4 years ago by patrick-kidger
  0

• Allow piecewise X and func not to have to use the same time pieces.

  #6 opened 4 years ago by patrick-kidger
  0

• Allow func-as-list and X-as-list

  #4 opened 4 years ago by patrick-kidger
  0

• Just leave computed parameters as buffers and it should now be fine?

  #3 opened 4 years ago by patrick-kidger
  0

• Update example using fixes from controldiffeq

  #1 opened 4 years ago by patrick-kidger
  0