Shapiro, A. and Niblett, T. (1982). Automatic induction of classification rules for a chess end-game. In M. Clarke (Ed.), Advances in Computer Chess 3 (pp. 73-92). Oxford: Pergamon.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....special purpose hardware. This model has been so successful that little else has been tried. The alternative AI approaches have not fared well due to the expense in applying the knowledge that had been supplied to the system. Those times in recent years that chess has been applied as a testbed [8, 27, 19, 21, 22, 30, 33, 26, 20] only a small sub domain of the game was used, so that fundamental efficiency issues that AI must grapple with have been largely unaddressed. However, we feel that there is a third approach that neither relies on search or the symbolic computation approach of knowledge oriented AI: what we shall ....

T Niblett and A. Shapiro. Automatic induction of classification rules for chess endgames. Technical Report MIP-R-129, Machine Intelligence Research Unit, University of Edinburgh, 1981.

....context rule takes as input a particular state and the sequence of all states in the last search and returns a pattern to be inserted into the database. A search context rule is a deterministic procedure that builds up a pattern given the previously mentioned inputs. In concept induction schemes [9, 11, 13, 16] the goal is to find a concept description to correctly classify a set of positive and negative examples. In general, the smaller description that does the job, the better. Sometimes the concept description needs to be made more specific to make a further distinction. At other times it can be ....

T Niblett and A. Shapiro. Automatic induction of classification rules for chess endgames. Technical Report MIP-R-129, Machine Intelligence Research Unit, University of Edinburgh, 1981.

....pattern to be inserted into the database. A search context rule is just a deterministic procedure that builds up a pattern given the previously mentioned inputs. Examples of search context rules can be found in Section 0.3.3. In concept induction schemes (Michalski, 1983; Mitchell et al. 1986a; Niblett and Shapiro, 1981; Quinlan, 1986) the goal is to find a concept description to correctly classify a set of positive and negative examples. In general, the smaller description that does the job, the better. Sometimes the concept description needs to be made more specific to 2.2. Adding and removing patterns 3 make ....

T Niblett and A. Shapiro. Automatic induction of classification rules for chess endgames. Technical Report MIP-R-129, Machine Intelligence Research Unit, University of Edinburgh, 1981.

....1979] When a clash occurs, i.e. when there are two positions P 1 and P 2 with different class values and 8P 2 TEST;P(P 1 ) P(P 2 ) then the two positions are given to the user who should produce a new predicate or recode an old predicate in order to discriminate the two positions. [Shapiro and Niblett, 1983] Figure 7: The ID3 algorithm [Quinlan, 1984, Utgoff, 1988] 3.2.1 Losing in 0 move positions The ID3 algorithm constructed the following tree to discriminate losing in 0 positions from winning or drawing positions(i.e. the result in losing in a least 1 move positions is not defined) KQ vs K ....

Shapiro, A. and Niblett, T. (1983). Automatic induction of classification rules for a chess endgame. In Clarke, M., editor, Advances in Computer Chess, volume 3, pages 73--92. Pergamon Press.

....and mutation operators as in genetic algorithms 3. goal and subgoal regression as in explanation based generalization. 4. node ordered induced subgraphs The proper mix of these methods is an important issue currently being explored. Generalization and Specialization In concept induction schemes [28,31,33,36] the goal is to find a concept description to correctly classify a set of positive and negative examples. In general, the smaller description that does the job, the better. Sometimes the concept description needs to be made more specific to make a further distinction whereas at other times it can ....

T Niblett and A. Shapiro. Automatic induction of classification rules for chess endgames. Technical Report MIP-R-129, Machine Intelligence Research Unit, University of Edinburgh, 1981.

....no results from this endeavor have been published. A severe problem with this and similar experiments was that, although the derived decision trees were shown to be correct and faster in classification than extensive search algorithms, they were also incomprehensible to chess experts. Shapiro [46] tried to alleviate this problem by decomposing it into a hierarchy of smaller sub problems that could be tackled independently. A set of rules was induced for each of the sub problems which together yielded a more understandable result. This process of structured induction has been employed to ....

Alen D. Shapiro and Tim Niblett. Automatic induction of classification rules for a chess endgame. In M. R. B. Clarke, editor, Advances in Computer Chess 3, pages 73--92. Pergamon, Oxford, 1982.

....However, no results from this endeavor have been published. A severe problem with this and similar experiments was that, although the derived decision trees were shown to be correct and faster in classification than extensive search algorithms, they were also incomprehensible to chess experts. [Shapiro and Niblett, 1982] tried to alleviate this problem by decomposing it into a hierarchy of smaller sub problems that could be tackled independently. A set of rules was induced for each of the sub problems which together yielded a more understandable result. This process of structured induction has been employed to ....

Alen D. Shapiro and Tim Niblett. Automatic induction of classification rules for a chess endgame. In M. R. B. Clarke, editor, Advances in Computer Chess 3, pages 73--92. Pergamon, Oxford, 1982.

....reported that when ID3 s output on the chess domain was shown to a domain expert, i.e. a chess master, it was completely opaque. Although it was very accurate, the tree was large, obscure, and the chess master was in a total blackout. The phenomenon was confirmed for related chess material in Shapiro Niblett (1982) and similar claims about other domains were made by Cendrowska (1987) Some researchers, and most of the Statistics community, use error rates (one minus the accuracy) instead of accuracy. 1 We chose to use accuracy because it is the more common 1 Jerry Friedman says that computer scientists ....

Shapiro, A. & Niblett, T. (1982), Automatic induction of classification rules for a chess endgame, in M. R. B. Clarke, ed., "Advances in Computer Chess", 3, Oxford: Pergamon.

....systems ID3 [15] and AQ11 [12] In order that the application of statistical measures is reliable, it is necessary to have a large number of training examples available. In many domains where the constraint for a large number of examples is met, rule induction methodology has proved successful ([18, 17, 11, 16]) However to extend this methodology to other domains, the ability to represent and exploit available domain knowledge in addition to that implicit in the example set is required. This is necessary in domains where statistical evidence alone is inadequate for reliably constraining the search for ....

A. Shapiro and T. Niblett. Automatic induction of classification rules for a chess endgame, pages 73-- 91. Volume 3, Pergamon, Oxford, 1982.

....tree was induced for the lost 2 ply problem, and a 177 node tree for the 3 ply problem. The application of these inductive techniques in larger domains is clearly likely to produce large trees which are usually difficult to comprehend. The motivation behind the method of structured induction [122] was to extend the inductive approach to produce humanly comprehensible decision trees, even for very large problems. This was achieved by having the expert decompose the problem into sub problems as is done in the methodology of structured programming. The ID3 algorithm was then used to induce a ....

A. Shapiro and T. Niblett. Automatic induction of classification rules for a chess end-game. In M. R. B. Clarke, editor, Advances in Computer Chess, volume 3, pages 73--92. Pergamon Press, Oxford, 1982.

....3. goal and subgoal regression as in explanationbased generalization. 4. node ordered induced subgraphs The proper mix of these methods is an important issue currently being explored. Generalization and Specialization In concept induction schemes [ Michalski, 1983; Mitchell et al. 1986; Niblett and Shapiro, 1981; Quinlan, 1986 ] the goal is to find a concept description to correctly classify a set of positive and negative examples. In general, the smaller description that does the job, the better. Sometimes the concept description needs to be made more specific to make a further distinction whereas at ....

T Niblett and A. Shapiro. Automatic induction of classification rules for chess endgames. Technical Report MIP-R-129, Machine Intelligence Research Unit, University of Edinburgh, 1981.

....proved valuable as tools for assisting in the task of knowledge acquisition for expert systems. In particular, two families of systems based on the ID3 and AQ algorithms have been especially successful. ID3 has been successfully applied to the classification of chess endgames (for example [1] [2], 3] which were intractable to human experts because of the volume of data, and C4 (an ID3 descendant) to the diagnosis of thyroid diseases [4] Systems based on the AQ algorithm, such as AQ11 [5] and GEM [6] have been successful in the fields of soya bean diagnosis ( 6] 7] and chess [6] We ....

Shapiro A., Niblett T. (1982) Automatic induction of classification rules for a chess endgame, Advances in Computer Chess 3. Ed. Clarke M.R. Oxford: Pergamon pp.73-91.

No context found.

Shapiro, A. and Niblett, T. (1982). Automatic induction of classification rules for a chess end-game. In M. Clarke (Ed.), Advances in Computer Chess 3 (pp. 73-92). Oxford: Pergamon.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC