diy-neural-network

Simple DIY neural network to understand backpropogation algorithm

Project Structure

This project is organized into two main directories:

• java/ - Java implementations using Maven multi-module structure
• python/ - Python implementations with modern packaging

Java Implementation

Modern ML module (Deeplearning4J)

Added a standalone Maven module in java/modern-ml that trains a tiny MLP to solve XOR using Deeplearning4J and ND4J.

Run it:

cd java/modern-ml
mvn -q -DskipTests exec:java

Main class: org.example.Deeplearning4JPerceptron

Build and run with Maven (multi-module)

This repo is now a Maven multi-module project with:

• java/legacy: builds the existing sources under Neural Network/src
• java/modern-ml: Deeplearning4J XOR example

Build everything from the java directory:

cd java
mvn -q -DskipTests package

Run legacy (original) app:

cd java/legacy
mvn -q -DskipTests exec:java

Run legacy but invoke modern DL4J example via Main argument:

cd java/legacy
mvn -q -DskipTests exec:java -Dexec.args=dl4j

Python Implementation

The Python directory contains scaffolding for neural network implementations. See python/README.md for more details.

To get started:

cd python
pip install -r requirements.txt