Including Ontologies in Monte-Carlo Tree Search and Applications – IOMCA

After having worked between 2004 and 2006 on various forms of stochastic dynamic programming techniques, OT is working since 2006 on Monte-Carlo Tree Search, a revolutionary tool for high-dimensional planning applied to the famous game of Go. OT has also worked on applications of MCTS to other fields of computer science (see technical context below) and is convinced that MCTS is definitely much more than the best current tool for computer-Go and should be used also in industrial applications and compared to classical tools in this context. AR has a long experience around planning in energy management and is the CEO of the company Artelys which is the author of several studies, formations, tools around planning in energy. OT and CSL have worked on the application to the game of Go. They have seen that huge improvements are possible thanks to knowledge management; huge databases of games are used for extracting patterns which introduce a helpful bias in MCTS. The next step is to use this kind of knowledge management for long term dependencies ? in the case of Go, this is related to solving ?semeais', a great challenge for many go programmers that we'd like to solve with general purposes and not ad hoc solvers. OT, AR and Artelys aim to collaborate around a comparative study of (i) standard tools for stochastic optimization (ii) more recent tools based on MCTS. OT and CSL will collaborate on the inclusion of knowledge management techniques for MCTS applied to the game of Go. The collaboration is aimed at producing methods based on automatic extraction, so that they can be used for a wide range of applications.