Spiking Network Simulation

This repository contains a Jupyter Notebook that simulates a spiking neural network (SNN) using PyTorch. The network is trained and tested on the Spiking Heidelberg Digits (SHD) dataset.

Files

• spiking-network.ipynb: Jupyter Notebook containing the code for setting up, training, and evaluating the spiking neural network.
• utils.py: Python utility file for downloading and preparing the SHD dataset.

There is also a folder labeled training_outputs where loss graphs and test accuracies for each hyperparameter configuration are saved. Files from a given execution must be moved out of this folder prior to performing a new execution; this will not be done automatically.

Environment Setup

To run the notebook, you need to set up a Python environment with the necessary dependencies. This project uses PyTorch, among other libraries.

Dependencies

• Python 3.8+
• PyTorch
• NumPy
• Matplotlib
• Seaborn
• h5py

You can install these dependencies via pip:

pip install numpy matplotlib seaborn h5py torch torchvision

Dataset

The SHD dataset will be automatically downloaded and prepared by the utils.py script when you run the notebook.

Execution Instructions

1. Clone this repository to your local machine.

2. Ensure that you have Jupyter Notebook or JupyterLab installed. If not, you can install it using:

   pip install jupyterlab

3. Navigate to the repository directory and start JupyterLab or Notebook:

   jupyter lab

   or

   jupyter notebook

4. Open spiking-network.ipynb in Jupyter and run the cells sequentially.

Notebook Structure

The notebook includes the following key components:

• Data Loading and Preparation: The notebook begins with loading the SHD dataset using utility functions defined in utils.py.
• Network Configuration: Parameters such as the number of neurons, learning rate, and other hyperparameters are set up.
• Model Definition: Definition of the spiking neural network model, including forward and backward passes.
• Training Loop: Code for training the SNN on the training dataset.
• Evaluation: Functions to evaluate the model's performance on the test set.
• Visualization: Code to visualize training loss and spike trains.

Ensure your Python environment is properly set up and all dependencies are installed before running the notebook to avoid runtime errors.