Spiking-Neural-Network-Simulation-for-Pattern-Recognition
Lakshitha A
Like this project
Posted Feb 12, 2025
Contribute to lakshiii/Spiking-Neural-Network-Simulation-for-Pattern-Recognition development by creating an account on GitHub.
Spiking-Neural-Network-Simulation-for-Pattern-Recognition
This repository implements a Spiking Neural Network (SNN) trained on the Fashion-MNIST dataset, which consists of 28x28 grayscale images of clothing items. The code uses Brian2, a Python library for simulating spiking neural networks, and leverages the Leaky Integrate-and-Fire (LIF) neuron model for simulation. DataSet: www.kaggle.com/datasets/zalando-research/fashionmnist?resource=download
Features
Fashion-MNIST dataset: This dataset contains 60,000 training images and 10,000 test images of fashion items, each represented as a 28x28 pixel grayscale image.
Poisson Spike Encoding: Converts images into spike trains using a Poisson process to represent pixel intensities.
Leaky Integrate-and-Fire (LIF) Neuron Model: Implements a simple LIF model for simulating excitatory neurons.
Synaptic Plasticity (STDP): Incorporates Spike-Timing-Dependent Plasticity (STDP) for synaptic weight adjustment based on the timing of spikes.
Training & Evaluation: Trains the network on batches of Fashion-MNIST images and evaluates the performance on a small test set.
Requirements
Python 3.x
Brian2: For simulating spiking neural networks.
NumPy: For numerical computations.
Pandas: For data loading and manipulation.
Matplotlib: For plotting spike raster.
To install the necessary libraries, run: