Madbits Tutorial

Machine Learning with Torch7

These tutorials should be read/done in order.

To follow these tutorials, and to the exercise, you will need to download the associated code, which is hosted on GitHub, on this page. In this repository, each directory corresponds to a tutorial here.


Machine Learning is a field that relies heavily on the use of standard datasets, to evaluate and compare algorithms.

In the GitHub repository, you will find a directory called A_datasets. This directory provides one script per dataset. Each script automatically downloads the dataset, and loads it into the Torch environment.

If you want to quickly browse a dataset, call the corresponding script like this:

th -lgfx.start   # start gfx server, navigate to http://localhost:8000
th -i mnist.lua
> trainData
{[data]   = ByteTensor - size: 60000x1x32x32
 [labels] = ByteTensor - size: 60000}
> testData
{[data]   = ByteTensor - size: 10000x1x32x32
 [labels] = ByteTensor - size: 10000}

For now, we provide code for these datasets:

Subpages (6): 1 2 3 4 5 6