https://github.com/bblodfon/digit-classifier
A Machine Learning project that uses EM and Bernoulli mixes to classify digits
https://github.com/bblodfon/digit-classifier
digit-classifier digit-recognition machine-learning matlab mnist
Last synced: about 1 month ago
JSON representation
A Machine Learning project that uses EM and Bernoulli mixes to classify digits
- Host: GitHub
- URL: https://github.com/bblodfon/digit-classifier
- Owner: bblodfon
- License: mit
- Created: 2015-02-17T14:16:05.000Z (over 11 years ago)
- Default Branch: master
- Last Pushed: 2020-02-15T18:27:41.000Z (over 6 years ago)
- Last Synced: 2025-09-09T14:08:14.774Z (11 months ago)
- Topics: digit-classifier, digit-recognition, machine-learning, matlab, mnist
- Language: MATLAB
- Homepage:
- Size: 12.5 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE.md
Awesome Lists containing this project
README
# 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](https://github.com/bblodfon/digit-classifier/blob/master/Digit%20classification%20(EM-BernoulliMix).pptx) you can find the presentation I did of this project for the Machine Learning Course, during my MSc (AUEB) in 2014.