sjb@cs.toronto.edu (Stephen Bellantoni) (09/30/89)
>From Evangelos Milios: > > Next Tuesday, October 3, three prominent computer scientists from the > Soviet Union will be visiting our department. Giorgi Adelson-Velsky > is the inventor of AVL trees, and a solid theoretician. This is his > first trip outside the USSR (he was "on ice" before glasnost). A > second guy in the group is Misha Donskoy, developer of the KAISA chess > program (world champ some time ago). > > I attach the abstracts describing the work of the Soviet visitors. > Please send me e-mail if you would like to spend some time with them. > I expect the visitors to arrive here in late morning. Misha Donskoy > will give a talk at 4-5pm on computer chess. > > Thank you, > Evangelos > > On search ordering under time pressure. > > G.M. Adelson-Velsky, V.L.Arlazarov (M. Donskoy, translator) > Institute for Systems Studies, Moscow, USSR > > In the talk the properties of probabilistic features of positions > and moves are considered. Using values of them one can make > decisions on the search order under given time restrictions on move > choice. The heuristic rules are proposed for the choice of next move > to be searched in the depth of the tree and for the intermediate re- > evaluations of heuristic values of game tree positions. > > DEFINING PROBABILISTIC PROPERTIES OF A GAME > > G.M. Adelson-Velsky (M. Donskoy, translator) > Institute for Systems Studies, Moscow, USSR > (work with V.P.Akimov) > > In the talk the game of two opponents with full information is considered and > ways to define probabilistic properties of the game and of the heuristic > estimations of its positions are investigated. > > There will be stated conditions upon which the distribution of types of > positions in large enough depth of a tree does not depend of the type of the > root position of the tree. The dependence of utmost distribution from the > probabilistic game feature values and from heuristic estimation function is > determined. Using these results it is possible to determine these > probabilistic feature values by the statistic investigation of the game tree. > > Computer Chess and Artificial Intelligence > > M.V. Donskoy > Institute for Systems Studies, Moscow, USSR > > Computer Chess for a long time used to be a classical AI > problem. Is it true now? Yes, because following vital AI problems > could be solved using Computer Chess: > - How to understand the program's behavior; > - How to transfer local knowledge to global knowledge base; > - How to plan the program's activity. > There are specific computer Chess problems. Some of them are > very difficult and still not solved. They will be listed in the talk. > Their connection with AI will be discussed. > > > Computer Chess and Applied Problems. > > M.V. Donskoy > Institute for Systems Studies, Moscow, USSR > > When one looks for the solution of an applied problem he > usually uses hierarchical search. How to carry out this search is very > important for several reasons. This problem is well known in the > literature on Branch-and-Bound Method. In the talk the connection > between the BB and the Alpha-Beta Method for Computer Chess will > be discussed and the possibility of different search reduction > methods will be shown. With this goal in mind the exact posing of a > search reduction problem will be given. Some other possibilities to > utilize Computer Chess methods will be discussed. > > > > > > >