Neural-Networks-from-Scratch

Implements Neural Network models from scratch without any specialized libraries. Only Numpy is used for implementing the models. Scikit-learn has been used in some places for generating or loading datasets.

Single Layer Perceptron (SLP):

Applies Single Layer Perceptron (SLP) Algorithm to classify:

1. 2 Classes Dataset (Banknote Dataset)
2. 3 Classes Dataset where each pairwise set of classes is linearly separable (Iris Dataset)
3. Synthetically generated 2 Classes Dataset

Generalized Multi Layer Perceptron (MLP):

Applies Generalized MLP (Any number of Hidden Layers can be added via user input) Algorithm along with Momentum factor to classify:

1. Iris Dataset
2. Handwritten Digits Dataset