Neural Network From Scratch

This repository contains a custom neural network implemented from scratch, supporting both regression and classification tasks.

Components

1. Dense Layer

• Fully connected layer.

2. ReLU Activation

• Introduces non-linearity to the dense layer.

3. Loss Functions

a. Mean Squared Error (MSE)

• Used for regression tasks.

b. Cross Entropy Loss

• Used for classification tasks.

Examples

• Example training on different common datasets (EMNIST etc.) are underway.