Neural Network for Handwritten Digits Classification

This project implements a neural network to classify handwritten digits (0-9) using the MNIST dataset. The model is built using TensorFlow/Keras and achieves high accuracy by preprocessing the data and optimizing the network architecture.

Project Overview

The model is designed as a feedforward neural network with multiple layers. The steps involved in this project include:

1. Data Preprocessing: Loading the MNIST dataset and normalizing the data.
2. Model Building: Creating a neural network with one input layer, one hidden layer with ReLU activation, and an output layer.
3. Model Training: Training the model using the training dataset.
4. Evaluation: Evaluating the model’s accuracy using the test dataset and confusion matrix.

Dataset: MNIST

• Source: MNIST Handwritten Digits Database
• Description: The MNIST dataset consists of 60,000 training images and 10,000 test images, where each image is a 28x28 grayscale representation of a digit (0-9).
• Labels: Each image is associated with a label (digit) from 0 to 9.
• Usage: The dataset is widely used as a benchmark for testing machine learning models on image classification tasks.

Technologies Used

• Python
• TensorFlow/Keras
• NumPy
• Matplotlib
• Seaborn

Steps

1. Data Loading & Preprocessing

The MNIST dataset is loaded and normalized using Z-Score normalization to improve accuracy.

2. Model Architecture

The model is designed with the following structure:

• Input Layer: Flattening the 28x28 pixel images into a 784-dimensional vector.
• Hidden Layer: A fully connected layer with Sigmoid / ReLU activation.
• Output Layer: 10 neurons corresponding to the digits 0-9, with a linear / softmax activation to output class probabilities.

3. Model Training

The model is trained using the Adam optimizer and Sparse Categorical Crossentropy loss function.

4. Evaluation

The accuracy of the model is evaluated using the test data, and the performance is visualized using a confusion matrix.

Results

• Accuracy: The model achieves an accuracy of 99.20% on the test dataset.
• Confusion Matrix: Provides detailed insights into correctly and incorrectly classified digits.

License

This project is licensed under the MIT License. See the LICENSE file for details.