Tutorial on Deep Learning and Unsupervised Feature Learning
Andrew Ng
Tutorial: Sparse Autoencoders
Here is the first of ten planned tutorial pieces on deep learning and unsupervised feature learning.
By completing this tutorial (i.e., reading the lecture notes and completing the programming assignment),
you will have implemented a sparse autoencoder, which is a basic unsupervised feature learning algorithm
useful for images/videos, audio, text, etc. (though this tutorial explains the ideas using images). This is
also a primary building block for more sophisticated deep learning algorithms.
You can also watch Lecture video 1 and Lecture video 2 on this material. (These videos were
made about a year ago and are on a slightly different version of the sparse
autoencoder than is described in these notes.)
Please feel free to contact Andrew Ng with comments/questions.
UPDATE: Additional tutorial materials are being put together,
that goes beyond this initial set of notes. You'll be able to learn
more about deep learning by going to this
new tutorial.