Measuring information transfer between neural spike trains

This repository contains the code used in my honours project at USYD, to quantify information transfer between neural spike trains.

The dataset used was published with this paper. We analysed Neuropixels data of spontaneuous activity in visual areas of the two mice Waksman and Krebs.

Information transfer is quantified using the information theoretic measure called 'transfer entropy'. The tool used for estimating transfer entropy between events (like neural spikes) in continuous-time was developed in this study.

We also applied the continuous-time estimator to infer effective networks of multivariate information transfers. You can read about these networks here.

Pairwise TE estimation

• pairwise_te.py contains transfer entropy estimations between pairs of cells in our dataset as implemented by jidt.
• association_pairwise_te.py contains the same method but was used to estimate TE between cells in different areas of the mouse brain in our dataset.
• plot_pairwise_te.py contains results post-processing and plotting.

Effective network inference

• eff_net_inf.py contains transfer entropy estimation inside an algorithm for effective network inference as implemented by jidt.
• plot_eff_nets.py contains results post-processing and plotting.

Pre-processing

• save_neuropixel_spike_times.m extracts relevant spike times from the Neuropixels dataset.
• prep_spikes_pickle.py processes spikes into python pickles for use in effective network inference.