Predictive coding in Pytorch

Overview

This repository houses the implementation of predictive coding used in the Bogacz Group. It offers an easy and opensource access to implementing a range of predictive coding networks.

In this repository you can also find tutorials that can be run on google colab for:

• predictive coding models for supervised learning
• Monte Carlo predictive coding models
• recurrent predictive coding models
• temporal predictive coding models

Structure of the repository

The repository includes:

• the predictive coding library under predictive_coding
• 1_supervised_learning_pc.ipynb, a tutorial on how to train a predictive coding model to perform classification on MNIST
• 2_unsupervised_learning_mcpc.ipynb, a tutorial on how to train a Monte Carlo predictive coding model on Gaussian data
• 3_memory_rpc.ipynb, a tutorial on how to train a recurrent predictive coding model on a memory task
• 4_sequential_memory_tpc.ipynb, a tutorial on how to train a temporal predictive coding model on a sequential memory task.

Usage

Follow the next steps to clone the code and setup the necessary python libraries:

git clone https://github.com/Bogacz-Group/PredictiveCoding.git
cd PredictiveCoding
pip install -r requirements.txt

Contact

For any inquiries or questions regarding the project, please feel free to contact us at [email protected] or [email protected].

Code Aknowledgements

This repository builds upon the following repositories/codes:

• https://github.com/YuhangSong/Prospective-Configuration