Multi-Layer Perceptron (MLP) from scratch to classify digits in the MNIST dataset
This project will implement a Multi-Layer Perceptron (MLP) from scratch to classify digits in the MNIST dataset. It's layers will have both forward and backward pass implementations. Inbuilt functions from PyTorch are not used, or other similar libraries for the implementation of the following.
Implementation
• Fully-connected layer with bias.
• ReLU and Sigmoid activations.
• Mini-batch Gradient Descent (SGD) and Cross-Entropy loss.
File is not uploaded as per "Eindhoven University of Technology" (TU/e) assignment guidelines.