It is demonstrated how the traditional genetic algorithm (GA) can be augmented by incorporating domain knowledge in the form of a version space (VS) into the structure. This hybrid inductive learning system is designed to handle problems in concept learning using the VS to control the search process that is performed by the GA. In this hybrid system a novel class of schemata is present called the hyperschema. A theorem for the hyperschema analogous to that for the traditional schema is presented. This theorem demonstrates how the addition of domain knowledge in the form of a VS allows the hybrid system to exploit schemata of higher order and defining length via a hitchhiking effect.<>
{"title":"The control of genetic algorithms using version spaces","authors":"R. Reynolds","doi":"10.1109/TAI.1990.130360","DOIUrl":"https://doi.org/10.1109/TAI.1990.130360","url":null,"abstract":"It is demonstrated how the traditional genetic algorithm (GA) can be augmented by incorporating domain knowledge in the form of a version space (VS) into the structure. This hybrid inductive learning system is designed to handle problems in concept learning using the VS to control the search process that is performed by the GA. In this hybrid system a novel class of schemata is present called the hyperschema. A theorem for the hyperschema analogous to that for the traditional schema is presented. This theorem demonstrates how the addition of domain knowledge in the form of a VS allows the hybrid system to exploit schemata of higher order and defining length via a hitchhiking effect.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123526294","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A discussion is given on how situation-driven autonomous systems can look ahead to resolve goal interactions and conflicts and to find a better way to achieve goals. Several lookahead schemes are examined implicitly in current autonomous systems. A rational reconstruction of a novel scheme in action selection dynamics is outlined for a better run-time planner to obtain goal-oriented behavior. It is shown that this reconstruction results in an action arbitration scheme having a more informative run-time lookahead for resolving goal interactions and conflicts without loss of its continual 'replanning' capability.<>
{"title":"Goal-oriented behavior in autonomous systems","authors":"Peiya Liu","doi":"10.1109/TAI.1990.130301","DOIUrl":"https://doi.org/10.1109/TAI.1990.130301","url":null,"abstract":"A discussion is given on how situation-driven autonomous systems can look ahead to resolve goal interactions and conflicts and to find a better way to achieve goals. Several lookahead schemes are examined implicitly in current autonomous systems. A rational reconstruction of a novel scheme in action selection dynamics is outlined for a better run-time planner to obtain goal-oriented behavior. It is shown that this reconstruction results in an action arbitration scheme having a more informative run-time lookahead for resolving goal interactions and conflicts without loss of its continual 'replanning' capability.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129458512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The problem of learning decision rules for sequential tasks is addressed, focusing on the problem of learning tactical plans from a simple flight simulator where a plane must avoid a missile. The learning method relies on the notion of competition and uses genetic algorithms to search the space of decision policies. In the research presented here, the use of available heuristic domain knowledge to initialize the population to produce better plans is investigated.<>
{"title":"Improving tactical plans with genetic algorithms","authors":"A. Schultz, J. Grefenstette","doi":"10.1109/TAI.1990.130358","DOIUrl":"https://doi.org/10.1109/TAI.1990.130358","url":null,"abstract":"The problem of learning decision rules for sequential tasks is addressed, focusing on the problem of learning tactical plans from a simple flight simulator where a plane must avoid a missile. The learning method relies on the notion of competition and uses genetic algorithms to search the space of decision policies. In the research presented here, the use of available heuristic domain knowledge to initialize the population to produce better plans is investigated.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"107 3-4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121001885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The ACKnowledge project aims at improving the efficiency of knowledge acquisition. For this purpose, the theoretical approach of the project consists of analyzing and evaluating the applicability of existing knowledge acquisition techniques in various contexts, in order to develop a comprehensive framework for integrating complementary knowledge acquisition techniques. Practical achievements of the project include the realization of a knowledge engineering workbench (KEW) based on the above framework. The authors introduce the overall approach of the project and present the central issue of integration within the project. The first prototype of KEW, which provides a realization of the theoretical approach, is presented.<>
{"title":"ACKnowledge project: A framework for knowledge acquisition techniques integration","authors":"F. Ramparany, M.-S. Doize, C. Jullien","doi":"10.1109/TAI.1990.130318","DOIUrl":"https://doi.org/10.1109/TAI.1990.130318","url":null,"abstract":"The ACKnowledge project aims at improving the efficiency of knowledge acquisition. For this purpose, the theoretical approach of the project consists of analyzing and evaluating the applicability of existing knowledge acquisition techniques in various contexts, in order to develop a comprehensive framework for integrating complementary knowledge acquisition techniques. Practical achievements of the project include the realization of a knowledge engineering workbench (KEW) based on the above framework. The authors introduce the overall approach of the project and present the central issue of integration within the project. The first prototype of KEW, which provides a realization of the theoretical approach, is presented.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117352468","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The authors describe BaLinda (Biddle and Linda) Lisp, a parallel execution Lisp dialect designed to take advantage of the architectural capabilities of Biddle (bidirectional data driven Lisp engine). The Future construct is used to initiate parallel execution threads, which may communicate through Linda-like commands operating on a tuple space. These features provide good support for parallel execution, and blend together well with notational consistency and simplicity. Unstructured task initiation and termination commands are avoided, while mandatory and speculative parallelisms (lazy versus eager executions) are both supported.<>
{"title":"BaLinda Lisp: a parallel list-processing language","authors":"C. Yuen, W. Wong","doi":"10.1109/TAI.1990.130409","DOIUrl":"https://doi.org/10.1109/TAI.1990.130409","url":null,"abstract":"The authors describe BaLinda (Biddle and Linda) Lisp, a parallel execution Lisp dialect designed to take advantage of the architectural capabilities of Biddle (bidirectional data driven Lisp engine). The Future construct is used to initiate parallel execution threads, which may communicate through Linda-like commands operating on a tuple space. These features provide good support for parallel execution, and blend together well with notational consistency and simplicity. Unstructured task initiation and termination commands are avoided, while mandatory and speculative parallelisms (lazy versus eager executions) are both supported.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131602429","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The author points out two important areas in which next-generation expert system shells could contribute to alleviate some of the problems associated with the maintenance of large expert systems: (1) supporting the development of a consistent domain model, and (2) improving the structure and the modularity of the rule base. The knowledge base of an expert system consists of a model of the problem domain and a description of the system's problem solving behavior. The maintainability of a knowledge-based system thus depends on a good design of both components. An example of a shell that exhibits some of the capabilities described here is CLASP.<>
{"title":"How could a shell improve the maintainability of expert systems?","authors":"J. Yen","doi":"10.1109/TAI.1990.130350","DOIUrl":"https://doi.org/10.1109/TAI.1990.130350","url":null,"abstract":"The author points out two important areas in which next-generation expert system shells could contribute to alleviate some of the problems associated with the maintenance of large expert systems: (1) supporting the development of a consistent domain model, and (2) improving the structure and the modularity of the rule base. The knowledge base of an expert system consists of a model of the problem domain and a description of the system's problem solving behavior. The maintainability of a knowledge-based system thus depends on a good design of both components. An example of a shell that exhibits some of the capabilities described here is CLASP.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130877684","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Most classic artificial-intelligence domains require satisfying a set of Boolean constraints. Real-world problems require finding a solution that meets a set of Boolean constraints and performs well on a set of real-valued constraints. In addition, most classic domains are static while domains from the real world change. In the present work, the authors demonstrate that SteppingStone, a general learning problem solver, is capable of solving problems with these characteristics. SteppingStone heuristically decomposes a problem into simpler subproblems, and then learns to deal with the interactions that arise between the subproblems. In lieu of an agreed-upon metric for problem difficulty, significant problems which are difficult for both people and programs are used as good candidates for evaluating progress. Consequently, the domain of logic synthesis from VLSI design is used to demonstrate SteppingStone's capabilities.<>
{"title":"Learning steppingstones for problem solving","authors":"D. Ruby, D. Kibler","doi":"10.1109/TAI.1990.130341","DOIUrl":"https://doi.org/10.1109/TAI.1990.130341","url":null,"abstract":"Most classic artificial-intelligence domains require satisfying a set of Boolean constraints. Real-world problems require finding a solution that meets a set of Boolean constraints and performs well on a set of real-valued constraints. In addition, most classic domains are static while domains from the real world change. In the present work, the authors demonstrate that SteppingStone, a general learning problem solver, is capable of solving problems with these characteristics. SteppingStone heuristically decomposes a problem into simpler subproblems, and then learns to deal with the interactions that arise between the subproblems. In lieu of an agreed-upon metric for problem difficulty, significant problems which are difficult for both people and programs are used as good candidates for evaluating progress. Consequently, the domain of logic synthesis from VLSI design is used to demonstrate SteppingStone's capabilities.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"174 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113967299","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
An application of genetic algorithms to the problem of discovering communication codes with properties useful for error corrections is described. Search spaces for these codes are so large as to rule out any exhaustive search strategy. Coding theory provides a rich and interesting domain for genetic algorithms. There are some coding problems about which a lot is known and good codes can be generated systematically. On the other hand, there are problem areas where little can be said about the characteristics of the codes in advance. Genetic algorithms have been advocated for these kinds of problems where domain knowledge is either limited or hard to represent and formalize. The authors describe some initial experiments on the use of genetic algorithms to discover maximal distance codes, and discuss the potential advantage of genetic algorithms in this problem domain.<>
{"title":"Discovery of maximal distance codes using genetic algorithms","authors":"K. Dontas, K. A. Jong","doi":"10.1109/TAI.1990.130442","DOIUrl":"https://doi.org/10.1109/TAI.1990.130442","url":null,"abstract":"An application of genetic algorithms to the problem of discovering communication codes with properties useful for error corrections is described. Search spaces for these codes are so large as to rule out any exhaustive search strategy. Coding theory provides a rich and interesting domain for genetic algorithms. There are some coding problems about which a lot is known and good codes can be generated systematically. On the other hand, there are problem areas where little can be said about the characteristics of the codes in advance. Genetic algorithms have been advocated for these kinds of problems where domain knowledge is either limited or hard to represent and formalize. The authors describe some initial experiments on the use of genetic algorithms to discover maximal distance codes, and discuss the potential advantage of genetic algorithms in this problem domain.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114225960","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Tetsuya Higuchi, T. Furuya, Ken'ichi Handa, A. Kokubu
The authors describe an associative parallel processor, IXM2, for artificial-intelligence applications, using network structured data (e.g., a semantic network). IXM2 consists of 64 associative processors and nine network processors, having a total of 256 Kwords by 40 bits of associative memory. IXM2 uses a novel interconnection mechanism based on complete connection, realizing both high communication bandwidth and expansibility into a larger system. It is shown that associative memory is very effective in suppressing the computational explosion in large semantic network processing. Arithmetic and logical operations can also be done in parallel on all the words of associative memory, for example, achieving 7200 MOPS for the 'less than' operation.<>
{"title":"IXM2: a parallel associative processor for semantic net processing-preliminary evaluation","authors":"Tetsuya Higuchi, T. Furuya, Ken'ichi Handa, A. Kokubu","doi":"10.1109/TAI.1990.130418","DOIUrl":"https://doi.org/10.1109/TAI.1990.130418","url":null,"abstract":"The authors describe an associative parallel processor, IXM2, for artificial-intelligence applications, using network structured data (e.g., a semantic network). IXM2 consists of 64 associative processors and nine network processors, having a total of 256 Kwords by 40 bits of associative memory. IXM2 uses a novel interconnection mechanism based on complete connection, realizing both high communication bandwidth and expansibility into a larger system. It is shown that associative memory is very effective in suppressing the computational explosion in large semantic network processing. Arithmetic and logical operations can also be done in parallel on all the words of associative memory, for example, achieving 7200 MOPS for the 'less than' operation.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"20 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115130883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The authors develop TCA*, a family of problem-independent, time-constrained, approximate A* search algorithms. The algorithms are designed to achieve the best ascertained degree of approximation with the minimum cost under a fixed time constraint, where cost is measured by either the cumulative space-time (CST) product or the maximum space encountered during the search. Only A* searches with admissible heuristic functions are considered; a branch-and-bound algorithm with a best-first strategy is an example of such a search. NP-hard combinatorial optimization problems whose feasible solutions can be found easily in polynomial time are studied. For the problems studied, it is observed that the execution time, the CST product and the maximum space required all increase exponentially as the degree of approximation decreases.<>
{"title":"TCA*-a time-constrained approximate A* search algorithm","authors":"B. Wah, Lon-Chan Chu","doi":"10.1109/TAI.1990.130355","DOIUrl":"https://doi.org/10.1109/TAI.1990.130355","url":null,"abstract":"The authors develop TCA*, a family of problem-independent, time-constrained, approximate A* search algorithms. The algorithms are designed to achieve the best ascertained degree of approximation with the minimum cost under a fixed time constraint, where cost is measured by either the cumulative space-time (CST) product or the maximum space encountered during the search. Only A* searches with admissible heuristic functions are considered; a branch-and-bound algorithm with a best-first strategy is an example of such a search. NP-hard combinatorial optimization problems whose feasible solutions can be found easily in polynomial time are studied. For the problems studied, it is observed that the execution time, the CST product and the maximum space required all increase exponentially as the degree of approximation decreases.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"91 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1990-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129576966","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：