Jacob Walker | 24 Apr 07:45 2012
Picon

GSOC Introduction - Implementing Gaussian Processes

Hi, if you have not yet heard from me on this mailing list/IRC, I am 
Jacob Walker
(also known as puffin444), and I have been selected to work on a project 
mentored by
Oliver Stegle for the Google Summer of Code. I am currently finishing up 
three
undergraduate degrees in Computer Science, Mathematics, and Economics at 
Michigan State University.
This fall I will be starting my Ph.D. at the Robotics Institute at 
Carnegie Mellon University.

In particular, I will be working on implementing Gaussian Processes in 
Shogun with an emphasis
on extendability and scalability to large datasets.

Here's an online book that describes GPs in detail 
(http://www.gaussianprocess.org/gpml/)

On a very general level, what I plan to do this summer is:

1. Implement the calculation of mean predictions and the covariance matrix
(This has practically already been done through the required application 
patch
submission).
2. Work on implementing learning of hyperparameters with the necessary 
code infrastructure.
3. Add sparse approximation, allowing the GPs to work on large datasets.
4. Add the ability to calculate evidence functions with non-gaussian 
likelihoods.

Much of what I intend to do has already been implemented in the MATLAB 
GPML toolbox
(http://www.gaussianprocess.org/gpml/code/matlab/doc/index.html); it 
should serve as a
great reference for testing.

I want to thank you all for giving me to opportunity to work on this 
great project, and again I want
to say that I am really excited to start working on this!

Sincerely,

Jacob Walker


Gmane