Applying the Forward-Forward Algorithm to the Event-Based Sensing

This is an implementation of the "Applying the Forward-Forward Algorithm to the Event-Based Sensing" paper.

Our paper: Accepted to The 7th International Conference on Machine Learning and Machine Intelligence (MLMI 2024)

Original Forward-Forward paper

The code base developed on the Implementation of Extending the Forward Forward Algorithm
Extending the Forward Forward Algorithm
Implementation of Extending the Forward Forward Algorithm

How to Run

To execute the code, run main.py.

The code performs tasks as defined in config.yml.

Pretrained Models and Configurations

Link to pretrained models

All the configurations YAMLs used in the paper are in the configs_models folder.

Testing Models

For testing pretrained models create a folder in configs_models/(MNIST|NMNIST) and copy your yaml and model.

After that, you need to change paths in test.py and execute it.

You can see results in the folder that you create as a .txt file.

How to Run on Google Colab

Step 1: Clone the Repository and Set Up the Environment

To start, clone the repository using the following command in the terminal. Set up the necessary libraries and login to Weights & Biases:

!git clone https://github.com/SuleymanEmirAkin/Event-Based_Forward-Forward.git
!pip install omegaconf
!pip install hydra-core --upgrade
!pip install wandb
!pip install tonic
import wandb
wandb.login(key=YOUR_KEY)

Step 2: Modify the Configuration

Adjust the configuration settings as necessary to fit your requirements.

Step 3: Run the Script

Finally, execute the main script with:

!python Event-Based_Forward-Forward/main.py