For best experience please turn on javascript and use a modern browser!

Almost graduated from the Master AI, student Igor Pejic is to present the first half of his thesis at GECCO, a highly regarded conference on optimization techniques.

Igor Pejic rehearsing his presentation for Gecco

"Our research was about solving perfect rectangle packing problems using Monte Carlo tree search, an algorithm widely successful in games such as go, chess, and even StarCraft. The perfect rectangle packing problem is an NP-complete problem with applications in the design of microprocessors and scheduling. Our results show that Monte Carlo tree search can solve some of these hard problems, and when it does solve them, it does that much faster and efficiently than an exhaustive algorithm.", says Igor.

"Our master students just seem to be getting better and better", says supervisor Daan van den Berg. "It is really remarkable that a student from Croatia, during lockdown in a foreign country, is still able to pull off such a performance. He's a real hard-hitter, I'm really proud to have such a student."

The conference, originally set for Cancun, Mexico, will be held completely online due to Covid-19.  

Their paper, "Monte Carlo Tree Search on Perfect Rectangle Packing Problem Instances" contains a digital tribute to Maarten van Someren (1955-2019). It is available upon request from Daan van den Berg.