CS/CNS/EE 156b: Learning Systems (Spring 2014)

Class Time: Tuesday and Thursday 2:30--3:55 pm in MRE 080.


- The server is now ready


- Presentations by the teams will start Tuesday, April 15, according the signed-up schedule in

080 Moore


- Here is the initial handout of the class.

- Here is a funny article that makes an excellent introduction to the project.


CS/CNS/EE 156 b. Learning Systems. 9 units (3-0-6); third term. Prerequisite: CS 156 a or equivalent. A project in machine learning based on the completed Netflix competition. Teams of two persons will carry out specific projects based on the Netflix data that will be provided for the class (no outside data is allowed). The project can explore different learning models and algorithms (including novel ones) as well as techniques for regularization and validation, aggregation, and optimization, and other aspects of machine learning that are applicable to this particular (huge) data set.


The Netflix competition, which took place from 2006 to 2009, was a fierce competition and left behind a trail of techniques and practical expertise that are worth studying. This link lists the reports of the winning team (three detailed reports that are linked to at the bottom in the first post). Also available locally in case the external link dies.

Here is a more recent article (part 1 and part 2) about the Netflix prize in retrospect and recommender systems in general.

Here is an introduction to Restricted Boltzmann Machines, one of the less documented techniques used in the competition.

The data is for use in this class only, and is not to be made available to others outside the class. You are asked to delete the data at the end of the term. Do not use data from any other source.

Submit your solutions to this server according to the instructions given in class. The scores and ranks will be posted on this board.

For background material in Learning Systems (CS 156a), the Learning From Data forum has been made accessible to you.

A Grand Prize of sorts :-) will be awarded to the best solution (on the test set), independently of the class grades.


e-mail Office Hours
Instructor Yaser Abu-Mostafa yaser(at)caltech... By appointment
TA Costis Sideris costis(at)caltech... --
TA Victor Kasatkin vikasatkin(at)gmail... --
TA Gregory Izatt gizatt(at)caltech... Mon, Wed 10-11 PM in Dabney (D47)
TA Bryan He bryanhe@caltech... --
Secretary Lucinda Acosta lucinda(at)caltech... --

Updated: 4/24/2014