/Neural-Networks-From-Scratch

A customized NeuralNetwork class for classification tasks using backpropogation and gradient descent

Primary LanguagePythonMIT LicenseMIT

Neural Networks From Scratch

Description:-

A customized NeuralNetwork class for classification tasks. The class also implements regularization and eases the application of regularization to the neural network. Dependencies include the numpy library. The following are the descriptions of the important function of the class.

gradientDescent:-

It performs Gradient Descent with iters as no. of iterations, alpha as the learning rate and lambda as regularization factor.

predict:-

It performs Forward Propogation on the neural network to classify the input.

getCostAndGradient:-

It performs Backward Propogation to calculate the gradients for each parameter