/Artificial-Neural-Networks-from-scratch-

Implementation of ANN(s) using basic python libraries

Primary LanguagePython

Implementation of Artificial Neural Networks

Libraries used

a. numpy
b. matplotlib

Description

Implemented ANN using basic numpy functionalities. The ANN model consists of following features
a. Mini-batch Gradient Descent.
b. L2 (ann_l2.py) and Dropout (ann_dropout) Regularization.
c. Input normalization.
d. Sigmoid, ReLU, Tanh, Linear activation functions and their derivative to be used while back propagation.
e. Variable dimensions of the network adjustable according to the user.
f. Initialized weights and biases in a manner to prevent vanishing and exploding gradients.
g. Cost function used is Mean Square Error.