Research internship at Microsoft Research in Cambridge, UK
Supervisor: Youssef Hamadi
Based on work by Kevin Leyton-Brown, we worked on using runtime prediction for automated parameter setting in search algorithms. We also employed Bayesian learning to enable incremental learning and a quantification of the uncertainty of our runtime estimates.
Here's a presentation I gave at the Cork Constraint Computation Center about this work.
TU Darmstadt
2004
M.Sc. thesis ("Diplomarbeit") at Darmstadt University of Technology in Darmstadt, Germany.
Title: "Stochastic Local Search for Solving the Most Probable Explanation Problem in Bayesian Networks"
Supervisors: Thomas Stützle, TUD & Holger Hoos, UBC
Here, I introduced new SLS algorithms for solving the MPE problem in arbitrary Graphical models. These algorithms scale very well for problems with high induced width, much better than state-of-the-art algorithms based on Branch-and-Bound with a Mini-Buckets heuristic.
Check out my webpage about MPE solving.
NASA Ames
Summers of 2002 and 2003
Summer Student Research Program (SSRP) -
Internship at NASA Ames Research Center
in Mountain View, California
Supervisor: Richard Dearden
We worked on Situational Awareness in Mars Rovers, applying Rao-Blackwellized Particle Filtering and Junction Trees for approximate online filtering in Dynamic Bayesian Networks.
Check out my webpage about the Gaussian Particle Filter.
UBC
2001/2002
Stochastic Local Search for computational design of RNA molecules that fold to a given Secondary Structure
Supervisors: Holger
Hoos, Anne Condon
Coworkers: Mirela
Andronescu, Anthony Fejes,
Firas Hamze
Check out the free RNA Designer Software at http://www.rnasoft.ca/
UBC
2001/2002
Development of the SAPS algorithm, a Dynamic Local Search algorithm for solving the propositional satisfiability problem (SAT)
Supervisor: Holger
Hoos
Coworker: Dave Tompkins
Check out the DLS for SAT webpage, maintained by Dave