Skip to content
master
Switch branches/tags
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
lib
 
 
 
 
 
 
 
 
 
 

README.md

The authors' official PyTorch SigCWGAN implementation.

This repository is the official implementation of [Conditional Sig-Wasserstein GANs for Time Series Generation]

Authors:

Paper Link:

Requirements

To setup the conda enviroment:

conda env create -f requirements.yml

Datasets

This repository contains implementations of synthetic and empirical datasets.

Baselines

We compare our SigCGAN with several baselines including: TimeGAN, RCGAN, GMMN(GAN with MMD). The baselines functions are in sig_lib/baselines.py

Training

To reproduce the numerical results in the paper, save weights and produce a training summaries, run the following line:

python train.py -use_cuda -total_steps 1000

Optionally drop the flag -use_cuda to run the experiments on CPU.

Evaluation

To evaluate models on different metrics and GPU, run:

python evaluate.py -use_cuda

As above, optionally drop the flag -use_cuda to run the evaluation on CPU.

Numerical Results

The numerical results will be saved in the 'numerical_results' folder during training process. Running evaluate.py will produce the 'summary.csv' files.

About

No description, website, or topics provided.

Resources

Releases

No releases published

Packages

No packages published

Languages