README.md

MNIST in Python with sklearn

The mnist dataset and description was obtained from here: http://yann.lecun.com/exdb/mnist/

This repository contains some utility functions for reading the MNIST data from the fileformat they come in.

There is a script create_image.py that is useful for visualizing the mnist data.

Example:

python create_image.py --input data/train-images.idx3-ubyte \
    --rows 10 --cols 10 \
    --output data\mnist_images_sm.png

This should output this image:

There is a script mnist_learning.py that uses sklearn to train a model, and score it's accuracy on the test data set.

Example:

python mnist_learning.py --train_images data/train-images.idx3-ubyte \
    --train_labels data/train-labels.idx1-ubyte \
    --test_images data/t10k-images.idx3-ubyte \
    --test_labels data/t10k-labels.idx1-ubyte

You can choose what classifier you want to use below the comment

###
### CHOOSE MODEL HERE
###

in the mnist_learning.py file.