This project focuses on developing a neural network capable of recognizing hand-written digits using the widely recognized MNIST dataset. The objective is to accurately classify these digits, which is also a well-known problem.
The initial approach involves utilizing a Multilayer Perceptron (MLP), a simple yet effective neural network architecture that demonstrates satisfactory performance for this particular problem.
Furthermore, an advanced Convolutional Neural Network (CNN) is employed as the second approach, emphasizing the progression towards deep layered neural networks. By leveraging the strengths of CNNs, we can achieve higher accuracy in image classification tasks like this.
This repository aims to conduct experiments and document the findings related to this well-known problem of digit recognition. The focus will be on comparing various aspects such as loss functions, activation functions, learning rates, and machine learning algorithms to analyze their impact on the model's performance. By exploring these different configurations, we can gain a deeper understanding of their influence on the accuracy and efficiency of the neural network.
"Neural Networks and Deep Learning. Michael Nielsen. (2017)"