Neural-network-made-in-java

This repo mostly serve as educational proposes, project itself is a implementation of one layer neural network (and can also be modified to be convulotional neural network). I made this solely for practicing and help other students understand the concept of machine learning with neural network. (No experience in outside library (numpy, etc.) needed)

Getting Started

First, bear with me through the mathematical model behind this project. It's crucial for understanding and using the library. To better understand back propagation and neural network, I recommend checking out 3Blue1Brown's video series. Feel free to skip this part if you are already familiar with the math behind neural network.

The Cost function

The partial derivative of weights and biases

Used Chain Rules to calculate the gradient of our neural network.

The derivatives used above

These are the derivatives we used to calculate the total gradient.

The definition of equations used in the neuron

the sigmoid function and how weights and biases are used.

After knowing the underlying mathematics, check out the documentation for how to use this project. Note: the neural network is not built with matrix, rather with simple for loop and nodes.

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

3Blue1Brown