Handwritten Digit Recognition using Neural Network

Overview

This project implements a neural network in C++ for recognizing handwritten digits. The neural network is trained to identify digits from 0 to 9 with an accuracy of approximately 96%. The project demonstrates the use of various C++ features such as classes, operator overloading, and exception handling.

Features

• Matrix Operations: Implements matrix creation, transposition, and various matrix manipulations.
• Activation Functions: Includes ReLU and Softmax activation functions.
• Layer Management: Supports creating and managing dense layers in the neural network.
• Digit Prediction: Predicts the digit in an input image using the trained neural network.

Getting Started

Prerequisites

• C++ compiler (e.g., g++)
• Makefile for building the project

Building the Project

To build the project, navigate to the project directory and run:

make mlpnetwork

Running the Program

The program expects binary files for weights and biases as input. To run the program, use the following command:

./mlpnetwork w1 w2 w3 w4 b1 b2 b3 b4

Once running, the program will prompt for the path to an image file containing a digit.

Example Usage

After providing the paths for weights and biases, the program will wait for an image file path. Upon providing the image file, the program will:

1. Load the image into a matrix.
2. Pass the matrix through the neural network.
3. Output the predicted digit and its probability.

Code Structure

• Matrix.h / Matrix.cpp: Implements the Matrix class with necessary operations.
• Activation.h / Activation.cpp: Implements activation functions (ReLU and Softmax).
• Dense.h / Dense.cpp: Implements a dense layer in the neural network.
• MlpNetwork.h / MlpNetwork.cpp: Implements the multi-layer perceptron network.

Example

./mlpnetwork path/to/weights/w1 path/to/weights/w2 path/to/weights/w3 path/to/weights/w4 path/to/biases/b1 path/to/biases/b2 path/to/biases/b3 path/to/biases/b4

Input an image file path when prompted:

path/to/image.png

The program will display the predicted digit and its probability.

Acknowledgements

Feel free to reach out with any questions or suggestions. Enjoy exploring the neural network implementation!