Building a Handwritten Digits Classifier with Neural Networks

by Nicholas Archambault

Deep neural networks have been used to reach state-of-the-art performance on image classification tasks in the last decade. For some image classification tasks, deep neural networks actually perform as well as or slightly better than the human benchmark.

This project explores the effectiveness of deep, feedforward neural networks at classifying images of digits.

Goals

1. Create data pipeline to smoothly implement machine learning algorithms and workflow, gradually building up more complex models.
2. Compare model accuracies of various model types, including:
   1. Baseline k-nearest neighbors model with k = 1 and four-fold cross-validation
   2. Single layer neural network with six different neuron quantities four-fold cross-validation
   3. Double layer neural network with three different neuron quantities four-fold cross-validation
   4. Triple layer neural network with three different neuron quantities and six-fold cross-validation

Output

Plots of model accuracy with varying parameters. Comparisons of characteristics of the most accurate models. An understanding of what contributes to a neural network model with a higher degree of accuracy.