SuyashSonawane / JavaNet

Understanding and Implementing Neural Networks in Java from Scratch 💻 Learning the popular concept in the most 💪strongly typed language

JavaNet

Understanding and Implementing Neural Networks in Java from Scratch 💻 Learning the popular concept in the most 💪strongly typed language
Explore the docs »

View Demo · Report Bug · Request Feature

About The Project

A simple Neural Network class completely written in Java from scratch without using any external libraries.

A Neural Network is a type of computational system which represents the human brain in a smaller manner. The Neurons are connected with Synapses, these concepts were taken from the 🧠. The system simulates the learning process of the brain by adjusting the Weights of the synapses by the process of Backpropagation.

Here's why you should consider using it:

• Easy to use and test
• Logging features
• Customizable

I also developed a even more feature rich JavaScript library which can be found here Lite Neural Network

Getting Started

1. Place NeuralNetwork.java in your project directory
2. Instantiate the neuralnetwork object
3. Process the data in proper format
4. Train the model
5. Test the model

For full explanation, please refer to the Article

Usage

Instantiating the neuralnetwork object

---

// In the sample we are creating a neural network with 2 input features and 1 output with 10 hidden nodes

// default constructor
NeuralNetwork nn = new NeuralNetwork(2, 10, 1); 

// constructor with custom learning rate
NeuralNetwork nn_custom_lr = new NeuralNetwork(2, 10, 1, 0.01); 

// constructor with multi-threading set to true
NeuralNetwork nn_with_multithreading = new NeuralNetwork(2, 10, 1, true); 

// constructor with multi-threading set to true and custom learning rate
NeuralNetwork nn_custom_lr_with_multithreading = new NeuralNetwork(2, 10, 1, 0.01, true); 

Data Format

---

// As we have 2 input features we created the 2d double array for X and 1d double array for Y

double[][] X = { { 0, 0 }, { 1, 0 }, { 0, 1 }, { 1, 1 } };
double[][] Y = { { 0 }, { 1 }, { 1 }, { 0 } };

Training the model

---

// Here we pass the data and specify the number of epochs

nn.fit(X, Y, 500); // silent learn

nn.fit(X,Y,500,0); // logging set to 0, shows training time and average error

nn.fit(X,Y,500,1); // logging set to 1, shows logs for each epoch

Testing the model

---

// Here we used 4 data-points to test the model, we iterate and pass each data-point to the model

List<Double> output;
double[][] input = { { 0, 0 }, { 0, 1 }, { 1, 0 }, { 1, 1 } };
for (double d[] : input) {
    output = nn.predict(d);
    System.out.println(output.toString());
}

Roadmap

See the open issues for a list of proposed features (and known issues).

Contributing

Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

Project status

• Basic Library
• Documentation
• Explanation Article
• Multi-Threading Support
• Interface to load data easily
• Multiple layer architecture
• Improve performance
• Unit Testing
• Production

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Your Name - @suyashysonawane - [e-mail]

Project Link: https://github.com/suyashsonawane/JavaNet