The Handwritten Digit Recognition is a project developed to classify digits drawn by the user in an interface.
he Training and Calculations of The Algorithm were implemented by Numpy Library.
The main objective of this project is to gain an in-depth understanding of how a neural network functions.
The Data Set : MNIST
The algorithm is a neural network with dense layers, following this configuration:
- Input Size: 784
- Hidden Layer 1: 50 neurons, activation function ReLU
- Hidden Layer 2: 20 neurons, activation function ReLU
- Output Layer: 10 neurons, activation function Softmax
Backpropagation: The batch gradient descent method was used with a batch size of 40 and 100 epochs, with a learning rate of 0.005.