Residual dynamics resolves recurrent contributions to neural computation

This repository contains all the neural and simulated datasets used in the above publication. All the data are stored as .mat files.

This repository is organized into three main folders. For ease of use, we recommend extracting the contents of each of the three folders into a single folder titled ‘data’.

neuraldata –Raw and preprocessed neural data. Organized into the following folders:

array_TDR_dotsTask_data – contains the non-processed neural data of each session for both monkeys

array_dotsTask_beh_xx – contains the behavioral data of each session for both monkeys (xx = monkey)

array_reppasdotsTask_datasegmented_binsize=xxms – contains the pre-processed neural data (session wise) of both monkeys for a specified bin size (xx = bin size)

simulations – Data from the simulated models. Organized into the following folders:

toymodels – Simulated data from all models of decisions/movement. Also includes simulations of the augmented line attractor and rotational dynamics models.

twoarearnn – Simulated data from the two types of (feedback and nofeedback) two-area RNN models, along with the impulse response simulations.

analyses – contains processed data files from various stages of the analysis pipeline. Only the most important files are listed below:

xx_aligned_reppasdotsTask_binsize=45ms.mat – session-aligned neural data for a specified monkey (xx = monkey)

xx_ndim=8_lag=3_alpha=200_50_allconfigsresdynresults.mat – contains the residual dynamics for a specified monkey.

xx_aligned_hankelagdimCV_reppasdotsTask_binsize=45ms.mat and xx_aligned_smoothnessCV_reppasdotsTask_binsize=45ms.mat – results of cross-validation of the residual dynamics pipeline for a specified monkey (xx = monkey)