Fully connected neural network in java

Implementation of multi-layer perceptron using basic Java(without any framework)

It is a simple implementation of complex concepts of gradient descent. I Used stochastic gradientdescent and binary cross-entropy loss function and applied it on MNIST dataset to achieve the accuracy ~88%. This project basically helps beginners with the understanding of neural networks. Interface is basic, but informative.

Structre of network:

Input layer -> Hidden layer -> Output layer

Dataset : http://yann.lecun.com/exdb/mnist/