For assignment 3, the code in mogEM.m appears to be correct. Do not use plotEllipse.m to visualize the Gaussians, as the Gaussians here are 64-dimensional, not two. You may need to play with the parameters (such as randConst) to get the 2 clusters to do something sensible. These notes, along with a couple other modifications can be found in the updated version of the assignment here. For example, the question of what to display in 1A has been clarified, as has the exploration strategy in 2.1.2b.
Also, the Q learning demo and code can be found here.
Finally, as I announced in class last Monday, the assignment is due at the beginning of tutorial on Friday.
There will be a Q&A session for the final exam, Monday December 14, 10-11 AM in Pratt 290C.
*********************************************************
Instructor: Richard Zemel, email: zemel@cs
Office: Pratt 290D, 978-7497
Office Hours: Thursday 4-5, or by appointment.
Lectures: MW 3-4 BA1220
Tutorials: F 3-4 BA1220
TA: Danny Tarlow
Further information for the course is available through these links: