/tgan-pytorch

A PyTorch implementation of Temporal Generative Adversarial Nets with Singular Value Clipping

Primary LanguagePythonMIT LicenseMIT

Experiment Setup

Experiment design for reference.

Baselines:

teporal GAN (this work)

scGAN (https://github.com/imsb-uke/scGAN): GAN-based simulator

Spaltter (https://github.com/Oshlack/splatter): Bayesian-based simulator

Synthetci Data:

To mimic existing time-series sc-RNAseq data, we generate a synthetic dataset with 10 time steps, 1000 cells, and 1000 genes.