Digital Recognition (MNIST)

Alice, Bob, and Daniel are friends learning machine learning together. After watching a few lectures, they are very proud of having learned many useful tools, including linear and logistic regression, non-linear features, regularization, and kernel tricks. To see how these methods can be used to solve a real life problem, they decide to get their hands dirty with the famous digit recognition problem using the MNIST (Mixed National Institute of Standards and Technology) database.

Hearing that you are an excellent student in the MITx machine learning class with solid understanding of the material and great coding ability in Python, they decide to invite you to their team and help them with implementing these different algorithms.

The MNIST database contains binary images of handwritten digits commonly used to train image processing systems. The digits were collected from among Census Bureau employees and high school students. The database contains 60,000 training digits and 10,000 testing digits, all of which have been size-normalized and centered in a fixed-size image of 28 × 28 pixels. Many methods have been tested with this dataset and in this project, you will get a chance to experiment with the task of classifying these images into the correct digit using some of the methods you have learned so far.

import mnist.utils as U
# Load MNIST data:
train_x, train_y, test_x, test_y = U.get_MNIST_data('mnist/Datasets/mnist.pkl.gz')
# Plot the first 20 images of the training set.
U.plot_images(train_x[0:20, :])