vikasnataraja / k-nearest-neighbors-for-mnist-digit-classification

This project explores the use of KNN algorithm to classify a database of handwritten digits. The algorithm does not use scikit-learn's built in KNNClassifier but instead a custom version that I have built.

K-Nearest Neighbors for Digit Classification

This project explores the use of KNN algorithm to classify a database of handwritten digits. The algorithm does not use scikit-learn's built in KNNClassifier but instead a custom version that I have built.

Getting Started

• This classifier uses helper functions like getCounts, majority, classify instead of the actual classifier.
• While the program does use scikit-learn for some parts, the main classifier is fully custom-built.

Prerequisites

• GitBash or any other shell
• Python
• Libraries including scikit-learn

To give you an example, I used GitBash for my shell and Python installed as well as Anaconda distribution (because it comes with most libraries or can be installed easily).

Installing and running the code

• To run the program, open KNN_Classifier.ipynb. This being a Jupyter notebook file, you can easily run the program on your local machine.
• Clone the repo to your local machine.
• Use GitBash or any other shell and navigate to the directory.
• Once in the directory, type jupyter notebook and you will be navigated to a browser window with the code.
• Once you see the code, feel free to implement the code, edit or anything else you might want to do.

Acknowledgements

A huge thank you to: Professor David Quigley at CU Boulder and stackoverflow for answering all my stupid questions