README.md

Retinal data used in "Inferring hidden structure in multilayered neural circuits"

This repository contains data used in the paper Inferring hidden data in multilayered neural circuits.

The data consists of the responses of 23 salamander retinal ganglion cells in response to 40 minutes of a spatiotemporal white noise stimulus.

These data were used to fit linear-nonlinear (LN) and multilayered linear-nonlinear (LN-LN) models to retinal data. The example scripts provided show how to fit these models on this data using the open source nems package.

Data

You can download the dataset needed to run the demo at this Google Drive link.

Demo

To run the demo, first install the nems package (and its dependencies). You also need to install numpy and the h5py package to load the data.

The demo can be run using: python demo.py

This will fit both an LN model and an LN-LN model to the same cell, without any regularization, and plot the learned model parameters for each. The nems package contains more information on how to save and test models and add regularization.

Example LN model parameters (linear filter and nonlinearity) that result from running the demo:

Citation

If you use the data in this repo, please cite our paper (https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006291)

@article{maheswaranathan2018inferring,
  title={Inferring hidden structure in multilayered neural circuits},
  author={Maheswaranathan, Niru and Kastner, David B and Baccus, Stephen A and Ganguli, Surya},
  journal={PLoS computational biology},
  volume={14},
  number={8},
  pages={e1006291},
  year={2018},
  publisher={Public Library of Science}
}