Handwritten Digit Classification

Supervised Learning of an RBF network to conduct Handwritten Digits Classification of MNIST dataset.

Overview

This repository contains a Jupyter Notebook that will train a RBF network on a real image dataset of handwritten number digits. The notebook includes the following:

• Robust Dataset:

  • 60,000 images in training set, 10,000 images in test set
  • Each image is 28x28 pixels
  • The images are grayscale (no RGB colors)
  • Each image (data sample) contains one of 10 numeric digit $0, 1, 2, \ldots, 8, 9$

• RBF network implementation without using an outside library

• 90+% Prediction Accuracy achieved on test set

Table of Contents

• Getting Started
• Installation
• Usage
• Project Structure
• Results
• Attribution

Getting Started

Prerequisites

• Python 3.8+ (or your preferred version)
• Jupyter Notebook (or JupyterLab)
• A list of Python libraries used in the notebook, for example:
  • pandas
  • numpy
  • matplotlib

Installation

1. Clone the repository to your local machine:

   git clone https://github.com/henrylanday/Handwritten-Digit-Classification.git
   

2. Navigate into the repository:

   cd Handwritten-Digit-Classification
   

3. Install required Python packages:

   pip install -r requirements.txt
   
   

Usage

1. Open the Jupyter Notebook (rbf_mnist.ipynb)

   jupyter notebook

2. Navigate to the notebook in the Jupyter interface and open it.
3. Run the notebook cells in order to reproduce the classifier

Project Structure

your-repo-name/
 ├── pycache/
 ├── classifier.py
 ├── data/
 │   ├── mnist_test_data.npy
 │   ├── mnist_test_labels.npy
 │   ├── mnist_train_data.npy
 │   ├── mnist_train_labels.npy
 │   ├── rbf_dev_test.csv
 │   ├── rbf_dev_train.csv
 │   └── Screenshot 20... (example screenshot file)
 ├── kmeans.py
 ├── rbf_mnist.ipynb
 ├── rbf_net.py
 ├── README.md
 └── requirements.txt

• pycache/: Auto-generated cache files for Python modules.
• classifier.py: Python module containing classification logic or classes/functions for classification tasks.
• data/: Contains datasets and supporting files:
  • mnist_test_data.npy, mnist_test_labels.npy: NumPy arrays for MNIST test samples and labels.
  • mnist_train_data.npy, mnist_train_labels.npy: NumPy arrays for MNIST training samples and labels.
  • rbf_dev_test.csv, rbf_dev_train.csv: Example CSV files used for development and testing.
  • Screenshot 20...: A screenshot file.
• kmeans.py: Implements the K-means clustering algorithm.
• rbf_mnist.ipynb: Main Jupyter Notebook performing analyses/experiments (e.g., training the RBF network on MNIST).
• rbf_net.py: Implements the RBF (Radial Basis Function) network.
• README.md: Project documentation file (this file).
• requirements.txt: A list of Python dependencies needed to run the project.

Results:

Created by Henry Landay