Convolutional Neural Network

A CNN written in Java from scratch

Key Features • Build Instructions • Screenshot • Credits • License

🧮 Key Features

• No external machine learning libraries used
• Handles dataset loading and training (MNIST-like format)
• Builds multiple layers dynamically
• Training loop with accuracy testing per epoch
• Console-based outputs for transparency and learning

🛠 Build Instructions

This project was developed using Visual Studio Code and Java 21.

To build it:

1. Clone the repo git clone https://github.com/fludar/neuralnetwork/
2. Follow this tutorial if you don't have Java already installed
3. Run the app

▶️ Usage

Example run from VS Code PowerShell terminal:

PS E:\github\neuralnetwork> & 'C:\Program Files\Eclipse Adoptium\jdk-21.0.4.7-hotspot\bin\java.exe' '-XX:+ShowCodeDetailsInExceptionMessages' '-cp' 'E:\github\neuralnetwork\bin' 'App'

🙏 Credits

• Convolutional Neural Network in Java from scratch

📖 What did I learn?

• How convolutional neural networks function internally
• How to implement layers and training logic manually in Java
• How to process dataset input for training and testing
• How backpropagation and accuracy calculation work step by step
• The challenges of implementing ML without external libraries
• Reinforced Java programming skills through a practical project