Task-specific invariant representation in auditory cortex

Heller, C.R., Hamersky, G.R., David, S.V.

This repository contains the code to run all analyses presented in Heller et al, 2023. The data is publicly available at: Dryad and should be downloaded as follows, prior to running any code in this repository:

Download data

• Navigate to our dataset which is hosted on Dryad.

• Click the dropdown "version files" button in the upper right corner of the page

• Click on elife2024.zip to download the data

• Extract the contents of this downloaded zip file into a new subdirectory of this repository called data

• Check that inside the eLife2024_Task repository you now have a subdirectory with the structure: data/elife2024/eLife_2024/<data_files>

• Note - You can extract the data to whichever location you want. However, if you put it in a different location than the one above, make sure to update the RESULTS_DIR variable in settings.py accordingly.

Create conda environment

In a terminal, cd into the cloned repository and run the following commmand:

conda env create -f environment.yml

This might take a minute or two to collect packages and solve the environment. This is normal. Once the process finished, this will create a conda environment called elife2024 which should have all the necessary dependencies installed for running the analysis in the repository.

Reproduce manuscript figures

Using the conda environment you just set up (e.g., conda activate elife2024), you should now be able to run the code that reproduces the manuscript figures. The scripts to produce each figure are located in a figure specific subdirectory of figure_scripts. For example, all scripts needed to generate the the contents of Figure 2 (and associated supplementals) are located in figure_scripts/fig2.

matplotlib backend

This can be specified in the settings.py while and should be used by all figure scripts. The default backend is QtAgg.

Recaching analysis files

Included in the data files downloaded from Dryad are a set of "analysis files" which contain cached results for analysis such as the pairwise stimulus decoding. These can all be regenerated, if desired, by re-running the analysis on your own machine. The scripts for re-caching all analyses are located in cacheScripts. At the top of each .py cache script you will find a brief description of the analysis that it will perform.