Earticle

Unlike other games such as chess, draughts and backgammon, computers are currently quite weak at the game of go ( baduk). Brute force is difficult due to the higher branching factor and game length. Human made algorithms become very complex before even approaching human strength on a subproblem of the game. One possible approach to this challenging problem is to use machine learning to let the program learn and improve without increased human effort. Machine learning has been successful in other games (e.g. draughts, backgammon). In this paper we give an overview of existing techniques. We discuss different aspects of learning, and propose some directions of research. In particular we believe that a first order representation language combined with a multistrategy learning system can achieve much more than what currently exists.

The study of expert behavior offers a unique perspective on cognition. By looking at extreme performance, cognitive scientists determine what parameters of the human information processing system can change with extensive practice, what parameters are relatively stable and what strategies can be used to overcome the limits imposed by these stable parameters. As expert must abide by the known limits of cognition, the information gained from them can be generalized to nonexperts, who, except perhaps for individual variations in the stable parameters, could become experts given enough practice and study in a domain. In this article, I will summarize some of the main studies in chess. I , to my regret, can not to review the findings on the baduk research, because there are few studies. It is expected that many good and creative works in a baduk domain will be produced in near future.

In Australia, there are three groups of Go players. The first group comprises the lst or 2nd generation immigrants from Korea, China or Japan, who learnt at an early age. The second group is a small number of Australian children and teenagers who discover the game through a relative, friend or teacher. The final group comprises Australians who discovered Go as adults, often at university. The author considers the training and retention of the third group as critical to the future of Go in Australia. This paper presents ten commonsense principles for the mentoring of weaker adult Go players by more experienced amateur players. These principles are based on adult educational research and on adult teaching and learning practices.

Amateurs have contributed a lot to Go as a Science. During the last decade their valuable activity has grown considerably because of the Internet. So time is ripe for a cross-section of their work. Many detailed studies can be found in various publications. Therefore this overview is broad rather than deep. It offers a wide range of new approaches and methods suitable for scientific study of Go. The discussed topics include Go rules mathematics, game trees, calculation size in computer Go, and functional languages. For the modern study field Go rules mathematics it is shown that Go can be well-defined as a finite game, how complex Go is, that a particular rule can nicely simplify study, and how a general result about cycles looks like. It is well known that game trees are used to represent reading in computer Go. However, they are flexible and also allow methodical analyses of endgame values, eyespace values, status types, or plays in pass fights. Computer Go suffers from exploding calculation sizes. Restrictions are possible in case of the examples ladders, capturing after a local series of threats, equal approach moves, or eye shapes. Finally, using joseki as an example functional languages formalize strategy by describing the meanings and intentions of single plays, sequences of plays, groups, joseki as a whole, and choices within joseki in contrast to playing elsewhere.

We present a method that enable to efficiently prove tactical theorems in the game of Go. We have experimented theorem proving with different tactical sub-games of the game of Go: the capture game, the connection game, the eye and the life and death games. Theorem proving works very well for tactical Go problem solving. The moves it finds are always correct and it finds the move faster than usual algorithms. In this paper we present our algorithms associated to concrete examples, and we outline the possible applications that might interest Go players: teaching programs for beginner's to learn to capture and connect stones, problem solving programs for intermediate players, and even some applications that might interest good players such as a perfect 5x5 Go problem solver and composer.

The world-wide web has given us an infra-structure which allows the smooth implementation of database driven technologies which use various contemporary techniques to enable desktop applications to fetch data from and submit data to centralized database servers located all over the world. We will outline the design and implementation of an innovative website which will enable both the professional and amateur Go player to profit from such a central repository based on Go games. The repository's kernel software engine is GOBASE, a Go database program which has been operational on the web since 1996 and is since then used by thousands of Go players to search for joseki, fuseki and other Go game properties from a collection of professional Go games. A new model is proposed for the repository, which will not use a static repository but will be completely dynamic. Users from all over the world can subscribe to the system and store their own Go games in the repository. Searches for game properties will from that moment on not only return professional games but their own submitted games as well. This way the Go player can compare his/her play with other Go players in a fully transparent way. The core data of the database. the collection of professional games. is maintained by either the professionals themselves, their national Go organizations or by a selected group of administrators qualified by the super-administrator to modify this data. Regular users can of course only modify their own subset of Go games. This model combines central services with the personal needs of the Go players world-wide without the need to install complicated software and huge amount of data on their desktop computers. WAP and UMTS-like services are likely candidates to extend these initial web services even further.