MNIST-Digit-Recognition

Table of Contents

• Project
  • Table of Contents
  • About The Project
  • File Structure
  • Logistic Regression
  • Neural Networks
  • Convolutional Neural Networks
  • Getting started
  • Contributors
  • Acknowledgements and Resources
  • License

About

Implementing Deep Learning for Handwritten Digit Recognition from scratch

File Structure

👨‍💻MNIST-Digit-Recognition
 ┣ 📂assets                            // Contains all the reference gifs, results images
 ┃ ┣ 📂results
 ┣ 📂data                              // MNIST dataset
 ┃ ┣ 📄mnist.pkl.gz
 ┣ 📂documentation                     // Documentation
 ┣ 📄neural_network.py
 ┣ 📄README.md

Neural Networks

Method 1: Initialing parameters with zeros

Results

Method 2: Initialing parameters randomly

Results

Method 3: Initialing parameters using He initialization

Results

Method 4: Implementing L2 regularization with He initialization

Results

Method 5: Implementing dropout regularization with He initialization

Results

Getting Started

Prerequisites

To download and use this code, the following python libraries are required:

• numpy
• matplotlib
• pickle
• gzip

Installation

Clone the project by typing the following command in your Terminal/CommandPrompt

git clone https://github.com/PritK99/MNIST-Digit-Recognition

Navigate to the MNIST-Digit-Recognition folder

cd MNIST-Digit-Recognition

Usage

Once the requirements are satisfied, use the following commands to run the respective python file:

python <name_of_the_file.py>

Contributors

• Prit Kanadiya

Acknowledgements and Resources

• Neural Network and Deep Learning by Michael Nielsen.
• Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization taught by Andrew NG, delivered through Coursera.

License

MIT License