Low-tensor-rank RNNs

This package allows fitting a low-tensor-rank recurrent neural network (ltrRNN) to neural data recorded over learning.

---

Installation

pip install ltrRNN

Examples

Quick example

import ltrRNN

# your_data is a numpy array of shape (trials, neurons, time)
# The dictionary of hyperparameters is described in the example notebook
ltrRNN.fit(hyperparameters, your_data)

# The real-time output of fitting can be found in the ./runs directory

Notebook

See the example notebook for an application of ltrRNN to simulated data.

Reference

A. Pellegrino_@, N. A. Cayco-Gaijc^†, A. Chadwick^†. (2024). Low Tensor Rank Learning of Neural Dynamics. https://proceedings.neurips.cc/paper_files/paper/2023/hash/27030ad2ec1d8f2c3847a64e382c30ca-Abstract-Conference.html.