HEPGAME is a research project that combines the world’s fastest computer algebra system for High Energy Physics equations, FORM, with insights from artificial intelligence.
The name combines High Energy Physics and Games.
If there is one thing that finding the Higgs has proven, it is that experiments are getting better and better. As the precision of High Energy Physics experiments increases, theory has to keep up. The precision of predictions from High Energy Physics equations has to improve. This typically involves large and complicated equations. FORM is a computer algebra system designed for high performance, able to solve very large physics equations, taking over some of the tedious manual labor of manipulating formulas by solving them automatically. The goal of HEPGAME is to improve the performance of FORM in a substantial way, by taking advantage of recent findings in game playing algorithms.
Ever since computer algebra systems were invented, there has been a deep connection between physics and artificial intelligence in the field of combinatorial search. The first computer algebra systems were created by physicists. Nowadays Matlab, Maple, and Mathematica are well-known tools for scientists worldwide.
Artificial intelligence has created many results where computers can solve problems where humans need to apply their intelligence. One of the most publicized results was the May 1997 match where Garry Kasparov was beaten by the chess-computer Deep Blue. The heart of all good chess computers is the minimax algorithm, that has been used successfully by researchers creating game playing programs for many different board games. However, for the oriental game of Go, minimax failed to produce strong programs. In 2006 a breakthrough occurred, when MCTS (short for Monte Carlo Tree Search) was introduced. MCTS has since been applied successfully to other domains as well. The aim of HEPGAME is to apply MCTS to FORM, since we have reason to believe it can be successful in this domain too.
After we had submitted our original grant proposal we found first evidence in support of the claim in that AI combinatorics can find better variable orderings for simplifying multicvariate polynomials with Horner’s rule, one of the classic mathematics algorithms. (See our publication list)
HEPGAME is led by Jos Vermaseren, of Nikhef’s Theory Group. Jos is the creator of the FORM Computer Algebra System. HEGAME is funded by an Advanced Grant of the European Research Council. The artificial intelligence part of HEPGAME is brought in by Jaap van den Herik and Aske Plaat who have a background in combinatorial game playing algorithms.