General: 663

General: 664

Projects

Home



P2: Spectral Clustering on Handwritten Digits Database

Author: Danielle Middlebrooks , Advisor: Kasso Okoudjou (Math)


Problem Statement Presentation

Project Proposal

Abstract
Spectral Clustering is a technique used to group together data points of similar behavior in order to analyze the overall data. The goal of this project will be to implement a spectral clustering algorithm on the MNIST handwritten digits database in which we will be able to cluster similar images using a similarity matrix derived from the dataset. We will develop code in order to implement each step of the algorithm and optimize to efficiently obtain a reasonable clustering of the dataset.



MidYear Progress Report and Presentation

Final Presentation , Final Report