Homework 3 (Programming part): Gaussian Process Regression

Note that this is only part of homework 3. See the course website and hw3.pdf.

Summary

You will implement Gaussian Process regression on noisy training data. The slides should be helpful. The provided code is here: hw3-code.tgz. You should only make changes in gp.m between

% TODO: START STUDENT CODE

and

% TODO: END STUDENT CODE

The training points are in (x, noisyY). The Gaussian Processes will be sampled and evaluated at the values in t. Your code should fill in pred and variance with the correct values for all the numbers in t (which is numTestPoints in size).

Expected Results

When executed with the default parameters (Executing gp();), you should obtain the following output:

Mean Standardized Log Loss:0.5074

Submission

You should submit your code (gp.m) by email to peter@cs.washington.edu. There is no writeup required for this question (but of course there is the additional MDP question as a part of HW3).