# Memristor_learning_model

**Repository Path**: hfutlc/Memristor_learning_model

## Basic Information

- **Project Name**: Memristor_learning_model
- **Description**: build Memristor model and simulate learning circuit by coding
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 1
- **Forks**: 0
- **Created**: 2021-03-29
- **Last Updated**: 2023-12-02

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Memristor learning model
Building memristor model and simulate learning circuit by python.
Inspired by the Spike learning method using STDP(Spike-Time dependency plasticity) material, which is memristor.

### Goal
In this github, we aimed to build a learning model to classify MNIST dataset by using Memristor model.
By composing memristors to a crossbar architecture, It is actually similar to the layer of Artificial Neural Network.

### SADP (with a reverse circuit)
Since memristance variation changes by the amplitude of input-spike,
model could learn things by using SADP(Spike-Amplitude dependency plasticity).
Besides, by simply adding a reverse-input memristor crossbar, the stability of the output got better.  

### Output
Tried out single epoch with 20000 MNIST data, 
and got about 50~60% accuracy.

### Memristance distribution after learning process
![alt Memristance distribution after learning process](https://raw.githubusercontent.com/flninja12/Memristor_learning_model/master/image.png)