/SNN_memristor_based

Spiking neural network simulator based on memrisor characteristics

Primary LanguagePython

Memristor crossbar & SNN simulator

Description

Spiking neural network simulator based on memristor characteristics and crossbar architecture

Table of Contents

Installation

If you whant to run some examples or construct your own SNN clone repo

git clone https://github.com/anddudkin/anddudkin_mem_project.git

Theory

Spiking neural network (SNN) operates with spikes. SNN takes spikes as input and produce output spikes with respect to learning rule.

We can use different types of neurons (IF,LIF, etc.). Dinamics of Leaky Integrate and Fire (LIF) neuron membrance presented in the picture below.

Implemented one of the basic SNN learning rules - STDP (Spike-timing dependent plasticity)

Weights plot from MNIST_example.py

Build_network

Baseline for constructing your own network

Crossbar architecture

Implemented crossbar takes into account impact of wire resistance

Also descrete conductance states were extracted from real memristive device and used in following simulations

License

How to Contribute

Contributor Covenant