Investigating Intra-Layer Neuronal Mechanisms

This project implements and investigates intra-layer neuronal mechanisms within spiking neural networks, such as Lateral Inhibition, k-Winners-Take-All (KWTA), and Homeostasis, using the CoNeX framework. The goal is to explore their effects on learning processes and network behavior in spiking neural networks trained with the Spike-Timing-Dependent Plasticity (STDP) learning rule.

Table of Contents

• Project Overview
• Implemented Features
• How to Run
• Results
• References

Project Overview

This project focuses on understanding the role of intra-layer neuronal mechanisms, such as Lateral Inhibition, KWTA, and Homeostasis, in shaping the behavior of spiking neural networks. By implementing these mechanisms and evaluating their effects on learning, we aim to enhance the differentiation of patterns and improve the learning processes within the network. The network architecture involves input and output layers, with input patterns encoded using Poisson encoding.

Implemented Features

1. Lateral Inhibition:

   • A mechanism that enhances pattern differentiation by inhibiting neighboring neurons in the output layer, making neurons more sensitive to spatially varying stimuli.

2. k-Winners-Take-All (KWTA):

   • A competitive mechanism that selects the top K neurons with the highest activity, enhancing the learning of distinct patterns in classification tasks.

3. Homeostasis:

   • A regulatory process that maintains equilibrium in neuronal activity by adjusting weights and neuron responses, ensuring stable learning and preventing network over-excitation.

4. STDP Learning Rule:

   • Spike-Timing-Dependent Plasticity is used to adjust synaptic weights based on the timing of spikes, reinforcing connections that contribute to a neuron’s firing while weakening less useful connections.

How to Run

1. Clone the repository:

   git clone https://github.com/MohaZamani/Intra-Layer-Neuronal-Mechanisms.git

2. Install the necessary dependencies:

   pip install -r requirements.txt

3. Run the simulation notebooks:

   • For Lateral Inhibition: Open and run lateral_inhibition.ipynb
   • For KWTA Mechanism: Open and run kwta.ipynb
   • For Homeostasis: Open and run homeostasis.ipynb

   You can launch the notebooks by executing:

   jupyter notebook

Results

Results from the simulations include:

• Weight Evolution: Plots showing how synaptic weights change in response to different input patterns and learning mechanisms.
• Cosine Similarity: A measure of similarity between weight vectors during learning, providing insight into the differentiation of patterns.
• Neuron Activity: Spike raster plots that visualize the activity of neurons across different experiments and mechanisms.

All simulation results and detailed analysis is available in the report.

References

• CoNeX Framework: GitHub Repository for CoNeX Framework
• STDP Learning: Spike-Timing-Dependent Plasticity on Wikipedia
• Homeostasis: Homeostasis in Neural Networks