pennatia

MNIST Modeling Project

Context

This project is an extension of the “Data Science Algorithms” repo that you can explore in the “repos” section. Essentially, it’s an exploration of the MNIST Datasets using various machine learning & data science methodologies, mostly built from scratch.

Scope & Methodology

The goal of this project was to:

  • Feature Reduce the MNIST Dataset
  • Understand which features and methodologies are most useful to accurately categorize a new input image
  • Generate a model, the analysis of which could be easily digested and understood.

The data science toolkit exemplified here is reduced, but it contains Z-Scoring, Mahalanobis Distance outlier removal, Parzen Windows, Linear Discriminant Analysis, as well as a Convolutional Neural Network.

.__======____=---=).................................................................
/\            \_--===)..............................................................
[ \ (0)   \~    \_-==)..............................................................
.\   \      / )J~~    \-=)..........................................................
..\  \\\\___/  )JJ~~~   \)..........................................................
...\_____/JJ~~~~~    \\.............................................................
.../ \  , \J~~~~~     \\............................................................
...(-\)\=|\\\\\~~~~       L__.......................................................
...(\\\\)  (\\\\\)_           \==__.................................................
...\V    \\\\\) ===_____   \\\\\\\\\\\\.............................................
...........\V)     \_) \\\\\\\\JJ\J\)...............................................
......................./J\JT\JJJJ)..................................................
.......................(UU)'...............................................@pennatia

Angelo Pennati

Occasionally known as the Carp Man.

2019-03-26