Skip to content

Latest commit

 

History

History
15 lines (10 loc) · 1.2 KB

README.md

File metadata and controls

15 lines (10 loc) · 1.2 KB

lfads-pytorch

PyTorch implementation of the LFADS architecture.

The Latent Factor Analysis via Dynamical Systems (LFADS) is a sequential autoencoder-based method designed with the purpose of inferring single-trial dynamics of neural activity. The major part of the model operates on neural firing rates, not spikes and can be applied to various data modalities.

For a basic application take a look at the toy example notebook.

The code in this repository is inspired by LFADS JAX implementation. Only the basic architecture is implemented. The controller is not yet a part of the repository.

A description of the method, its parameters and additional literature can be found in the above hyperlinks as well as here.

Contents

  • lfads.py - LFADS implementation.
  • lfads-toy-example.ipynb - a notebook with an easy to follow toy example.
  • utils.py - two functions supporting the notebook code.