Constrained predictive coding

This is the code used to generate the results from Constrained Predictive Coding as a Biologically Plausible Model of the Cortical Hierarchy. It runs on Python 3.9. To install, run the following in the repo folder:

conda env create -f conda.yml
conda activate cpcn
pip install .

You can instead use

pip install -e .

to create an editable install – useful for development.

The submit_task_* scripts in draft/hyper run the hyperoptimizations. These are written for the Flatiron Institute's SLURM system, and use disBatch. This is equivalent to running each line of the corresponding task_file_*.db.

After the hyperoptimization runs are completed, similar submit_task_* scripts are available in draft/simulations. These generate the data needed for making the figures.

Figures are generated using the various make_*_figures_*.py scripts in draft. Some exploratory code can be found in sandbox.