ajaypal91 / neural-network-from-scratch

Three implementations of Neural Network with back-propagation are demonstrated in the repository

Neural-Network-From-Scratch

Three implementations of Neural Network with back-propagation are demonstrated in the repository

There are three different implementations of Neural Network from scratch which uses backpropagation and gradient Descent.

DATASET USED: The dataset used is moons dataset provided in sklear library of python

Part 1) Contains mini-batch gradient descent

Part 2) uses the annealing schedule to change the value of epsilon

Part 3) uses sigmoidal activation function istead of tanh which is used in other examples.

The loss function is printed at each iteration and in the end the graph is plotted for model build on the moons dataset.

Run Test.py for running each example.