Skip to content

A Machine Learning project that uses EM and Bernoulli mixes to classify digits

License

Notifications You must be signed in to change notification settings

bblodfon/digit-classifier

Repository files navigation

digit-classifier

This Machine Learning project implements the Expectation Maximization algorithm using Bernoulli mixes (1-32 mixes were tested) in order to classify digits.

Run

Firstly, a train + test dataset is given on the mnist_all.mat file.

By running bernoullimix3.m file you train the dataset for all bernoulli mixes. If you run draw_digits.m after that for different values of Kplot (number of mixes), you will see what we have learned from the test dataset for each mix (larger K means a larger pool of different forms for every digit). Lastly, if you run the test_digits.m file you will get the final results (avg error percentage for every mix and for every digit/mix) on a results.txt file.

Install

The code was executed in Matlab (version: R2011a).

Presentation

Here you can find the presentation I did of this project for the Machine Learning Course, during my MSc (AUEB) in 2014.

About

A Machine Learning project that uses EM and Bernoulli mixes to classify digits

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages