About Me
I'm a PhD candidate in the
Machine Learning Group
of the
Computational and Biological Learning Lab
at the Department of Engineering in the University of Cambridge where my advisor is
Carl Edward Rasmussen;
my supervisor is
Zoubin Ghahramani.
I am supported by a
DataPath PhD Scholarship.
I have an undergraduate degree from
UC Santa Cruz.
At Cambridge, I am a member of
Magdalene College.
Research
My research interests are centered around probabilistic methods for machine learning. I deal with methods for various types of time series data: point processes, real valued data, and interval processes. I apply Gaussian processes, inhomogeneous Poisson point processes, and nonparametric tests to such data. I am interested in software engineering for machine learning and diagnostics that can be used to ensure MCMC, Gibbs Samplers, Variational Bayes, etc implementations are working; and how to debug them when they don't.
About My Sponsored Project
My research goals focus on applying machine learning, a combination of computer science and statistics, to predict failures in network hardware, particularly satellite Earth terminals. This task is known as condition monitoring. Software systems exist that record events on devices in satellite terminals, such as changes in signal strength or device temperatures. My research will give us the ability to make quantitative predictions on the time until the failure of each device. It is representable in much the same way as life tables are in the insurance industry. However, the predictions by my methods can change instantaneously in response to new events, which will provide more accurate predictions.
more
Snippet from First Year Report (Section 2.1)
DataPath specializes in monitoring and control software and satellite earth terminal units.
Its software, MaxView, is used for mission-critical applications in network management.
It is usually used to monitor satellite earth terminals and log the results of various
measurements. MaxView presents users with a virtual control board for monitoring
on a computer. It is designed so operators with minimal programming experience can
customize the GUI, automate certain network tasks, and write drivers that interface the
software to new network hardware. The basic premise of the DataPath problem is to
add a feature to the MaxView software that can give operators a warning when a fault is
likely to occur. Adding some prediction capability could potentially serve as value added
to the software. This type of problem falls in an area known as condition monitoring. It
also involves novelty detection because many of the failures may be rare.
more
Additional Info
I can be found at either my lab in Cambridge, my hometown of Santa Cruz, CA, or in Whistler, BC (especially around NIPS time). I frequent surfing (Santa Cruz), skiing/snowboarding/rafting (Whistler), and rowing in the Magdalene slow boat (Cambridge). I am also a dog and bear enthusiast.
My auxilary roles in the lab involve making frequent updates to the wiki and providing MATLAB utilities (and a few LaTeX ones too) in my svn Util directory. I also put pressure on my colleagues to right more readable and testable code.