The authors present results of experiments comparing the performance of the ID3 symbolic learning algorithm with a higher-order neural network (HONN) in the distortion invariant object recognition domain. In this domain, the classification algorithm needs to be able to distinguish between two objects regardless of their position in the input field, their in-plane rotation, or their scale. It is shown that HONNs are superior to ID3 with respect to recognition accuracy, whereas, on a sequential machine, ID3 classifies examples faster once trained. A further advantage of HONNs is the small training set required. HONNs can be trained on just one view of each object, whereas ID3 needs an exhaustive training set.<>
{"title":"An empirical comparison of ID3 and HONNs for distortion invariant object recognition","authors":"L. Spirkovska, M. B. Reid","doi":"10.1109/TAI.1990.130402","DOIUrl":"https://doi.org/10.1109/TAI.1990.130402","url":null,"abstract":"The authors present results of experiments comparing the performance of the ID3 symbolic learning algorithm with a higher-order neural network (HONN) in the distortion invariant object recognition domain. In this domain, the classification algorithm needs to be able to distinguish between two objects regardless of their position in the input field, their in-plane rotation, or their scale. It is shown that HONNs are superior to ID3 with respect to recognition accuracy, whereas, on a sequential machine, ID3 classifies examples faster once trained. A further advantage of HONNs is the small training set required. HONNs can be trained on just one view of each object, whereas ID3 needs an exhaustive training set.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"46 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":"128676597","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}
Genetic algorithms are a relatively new paradigm for search in artificial intelligence. It is shown that, for certain kinds of search problems, called permutation problems, the ordinary rule for intermixing the genes between two organisms leads to longer search chains than are necessary. A schema is a partially completed organism. Its order is the number of fixed components and its length is the distance between its first and last fixed component. A scheme is compact if its length and order are nearly equal. It is shown that the survival rate of a compact schema is directly proportional to the quality of the solution after a fixed number of iterations. The ordinary gene intermixing method called a crossover rule, separates the parents of a new organism at almost the precise point at which the compact scheme survival rate is at a minimum. A variation of the crossover rule is proposed that takes advantage of the knowledge of survival rates on the quality of the solution.<>
{"title":"Schema survival rates and heuristic search in genetic algorithms","authors":"B. Buckles, F. Petry, R. Kuester","doi":"10.1109/TAI.1990.130357","DOIUrl":"https://doi.org/10.1109/TAI.1990.130357","url":null,"abstract":"Genetic algorithms are a relatively new paradigm for search in artificial intelligence. It is shown that, for certain kinds of search problems, called permutation problems, the ordinary rule for intermixing the genes between two organisms leads to longer search chains than are necessary. A schema is a partially completed organism. Its order is the number of fixed components and its length is the distance between its first and last fixed component. A scheme is compact if its length and order are nearly equal. It is shown that the survival rate of a compact schema is directly proportional to the quality of the solution after a fixed number of iterations. The ordinary gene intermixing method called a crossover rule, separates the parents of a new organism at almost the precise point at which the compact scheme survival rate is at a minimum. A variation of the crossover rule is proposed that takes advantage of the knowledge of survival rates on the quality of the solution.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"2 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":"130297830","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 presents a method to check whether an implicit representation (i.e., a formula of the form t/(t1,. . .,tn), where t is viewed as a generalization of a set of examples and t1,. . .,tn are counterexamples) is a generalization with respect to a finite set of equations which describes the background knowledge problem; that is, whether there exists a ground (variable-free) instance of t which is not equivalent to any ground instance of t1,. . .,tn with respect to a set E of equations. Intuitively, the implicit representation t/(t1,. . .,tn) is a generalization if the set of ground instances of the formula t/(t1,. . .,tn) is non-empty. Whereas this problem is in general undecidable since the equality is so, it is shown here that, in the case where the set E of equations is compiled into a ground convergent term rewriting system, one can easily discover concepts in theories described by a finite set of equations.<>
{"title":"Learning from examples and counterexamples with equational background knowledge","authors":"Emmanuel Kounalis","doi":"10.1109/TAI.1990.130343","DOIUrl":"https://doi.org/10.1109/TAI.1990.130343","url":null,"abstract":"The author presents a method to check whether an implicit representation (i.e., a formula of the form t/(t1,. . .,tn), where t is viewed as a generalization of a set of examples and t1,. . .,tn are counterexamples) is a generalization with respect to a finite set of equations which describes the background knowledge problem; that is, whether there exists a ground (variable-free) instance of t which is not equivalent to any ground instance of t1,. . .,tn with respect to a set E of equations. Intuitively, the implicit representation t/(t1,. . .,tn) is a generalization if the set of ground instances of the formula t/(t1,. . .,tn) is non-empty. Whereas this problem is in general undecidable since the equality is so, it is shown here that, in the case where the set E of equations is compiled into a ground convergent term rewriting system, one can easily discover concepts in theories described by a finite set of equations.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"130 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":"116263282","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}
It is argued that constraint logic programming (CLP) systems can be built 'loosely' by separating a system into independent parts to avoid some repeated processing, and thus to increase further the efficiency of CLP systems. An architecture scheme is presented. Systems based on the architecture have desirable modularity. The system Conslog is described as an instance of the scheme. Conslog consists of two independent parts: a partial evaluator to carry out logical inferences and a solver in finite domains. The efficiency and behavior of the system are shown on a cryptarithmetic problem. The system is extensible.<>
{"title":"Combining partial evaluation and constraint solving: a new approach to constraint logic programming","authors":"Jin-Kao Hao, J. Chabrier","doi":"10.1109/TAI.1990.130387","DOIUrl":"https://doi.org/10.1109/TAI.1990.130387","url":null,"abstract":"It is argued that constraint logic programming (CLP) systems can be built 'loosely' by separating a system into independent parts to avoid some repeated processing, and thus to increase further the efficiency of CLP systems. An architecture scheme is presented. Systems based on the architecture have desirable modularity. The system Conslog is described as an instance of the scheme. Conslog consists of two independent parts: a partial evaluator to carry out logical inferences and a solver in finite domains. The efficiency and behavior of the system are shown on a cryptarithmetic problem. The system is extensible.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"1234 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":"115830985","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 present an extended production system architecture which can deal with forward reasoning in multiple contexts. The proposed architecture consists of a compiler of clauses and defaults into a Rete-like network, a Rete-based inference engine, and an assumption-based truth maintenance system (ATMS). The inference engine gives intermediate justifications to the ATMS and stores intermediate dependent assumptions of two-input nodes in the Rete-like network, allowing faster multiple-context reasoning. By means of this method, the multiple-context reasoner called APRICOT/0 has been implemented. An experiment under the logic design knowledge base shows that APRICOT/0 is about six to ten times faster than a system with a simple combination of a production system and the ATMS.<>
{"title":"A forward-chaining multiple-context reasoner and its application to logic design","authors":"Y. Ohta, Katsumi Inoue","doi":"10.1109/TAI.1990.130368","DOIUrl":"https://doi.org/10.1109/TAI.1990.130368","url":null,"abstract":"The authors present an extended production system architecture which can deal with forward reasoning in multiple contexts. The proposed architecture consists of a compiler of clauses and defaults into a Rete-like network, a Rete-based inference engine, and an assumption-based truth maintenance system (ATMS). The inference engine gives intermediate justifications to the ATMS and stores intermediate dependent assumptions of two-input nodes in the Rete-like network, allowing faster multiple-context reasoning. By means of this method, the multiple-context reasoner called APRICOT/0 has been implemented. An experiment under the logic design knowledge base shows that APRICOT/0 is about six to ten times faster than a system with a simple combination of a production system and the ATMS.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"32 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":"122282322","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 generic algorithm is presented, suitable for processing an important class of recursive queries, the so-called instantiated transitive-closure (TC) queries. The most important characteristics of this algorithm is that it reads any data-page from the system's disk at most once, i.e., the worst-case I/O behavior of this algorithm is linear with respect to the number of pages storing the database. Several variants, each with a different main-memory requirement, to the super-TC algorithm as well as a comparative performance evaluation of these variants are presented. The super-TC variant with minimum main-memory requirement is determined. A comparison of this variant with the more traditional delta -wavefront algorithm reveals the superiority of the super-TC variant, which is up to seven times faster.<>
{"title":"Super-TC: an efficient generic algorithm for processing the instantiated transitive-closure queries in deductive database systems","authors":"G. .. Qadah","doi":"10.1109/TAI.1990.130447","DOIUrl":"https://doi.org/10.1109/TAI.1990.130447","url":null,"abstract":"A generic algorithm is presented, suitable for processing an important class of recursive queries, the so-called instantiated transitive-closure (TC) queries. The most important characteristics of this algorithm is that it reads any data-page from the system's disk at most once, i.e., the worst-case I/O behavior of this algorithm is linear with respect to the number of pages storing the database. Several variants, each with a different main-memory requirement, to the super-TC algorithm as well as a comparative performance evaluation of these variants are presented. The super-TC variant with minimum main-memory requirement is determined. A comparison of this variant with the more traditional delta -wavefront algorithm reveals the superiority of the super-TC variant, which is up to seven times faster.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"6 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":"128522302","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 discuss high-level reasoning in a cooperative knowledge system and provide an effective high-level reasoning method, i.e. the multidimensional argument method. Based on the results of low-level reasoning this method of high-level reasoning can improve the accuracy of the whole cooperative system's results by using various kinds of information, such as mutual-evaluation information, self-evaluation, external knowledge and information, and dependence value. The authors also describe the development of the argument system PAT-1 which is a subsystem of DAI-tools HEDMUKCS.<>
{"title":"High-level reasoning in cooperative knowledge system","authors":"Fangqing Zhang, Da-you Liu, Rikun Tang","doi":"10.1109/TAI.1990.130338","DOIUrl":"https://doi.org/10.1109/TAI.1990.130338","url":null,"abstract":"The authors discuss high-level reasoning in a cooperative knowledge system and provide an effective high-level reasoning method, i.e. the multidimensional argument method. Based on the results of low-level reasoning this method of high-level reasoning can improve the accuracy of the whole cooperative system's results by using various kinds of information, such as mutual-evaluation information, self-evaluation, external knowledge and information, and dependence value. The authors also describe the development of the argument system PAT-1 which is a subsystem of DAI-tools HEDMUKCS.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"25 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":"130631873","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 the genetic programming paradigm, which genetically breeds populations of computer programs to solve problems. In genetic programming, the individuals in the population are hierarchical computer programs of various sizes and shapes. Applications to three problems in artificial intelligence are presented. The first problem involves genetically breeding a population of computer programs to allow an 'artificial ant' to traverse an irregular trail. The second problem involves genetically breeding a minimax control strategy in a different game with an independently acting pursuer and evader. The third problem involves genetically breeding a minimax strategy for a player of a simple discrete two-person game represented by a game tree in extensive form.<>
{"title":"Genetically breeding populations of computer programs to solve problems in artificial intelligence","authors":"J. Koza","doi":"10.1109/TAI.1990.130444","DOIUrl":"https://doi.org/10.1109/TAI.1990.130444","url":null,"abstract":"The authors describe the genetic programming paradigm, which genetically breeds populations of computer programs to solve problems. In genetic programming, the individuals in the population are hierarchical computer programs of various sizes and shapes. Applications to three problems in artificial intelligence are presented. The first problem involves genetically breeding a population of computer programs to allow an 'artificial ant' to traverse an irregular trail. The second problem involves genetically breeding a minimax control strategy in a different game with an independently acting pursuer and evader. The third problem involves genetically breeding a minimax strategy for a player of a simple discrete two-person game represented by a game tree in extensive form.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"134 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":"133947808","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}
K. Heisler, P. Johnson, W. Tsai, Y. Kasho, J. R. Snyder
Research methodology from cognitive science and artificial intelligence is applied to the problem of software representation to support the process of software change. A model of software to support change, suggested by the current research, is presented. Populating that model requires data from tasks related to code-based functional decomposition, data abstraction, and functional data flows. A competence model for the task of code-based functional decomposition is proposed.<>
{"title":"Software representation to support change","authors":"K. Heisler, P. Johnson, W. Tsai, Y. Kasho, J. R. Snyder","doi":"10.1109/TAI.1990.130405","DOIUrl":"https://doi.org/10.1109/TAI.1990.130405","url":null,"abstract":"Research methodology from cognitive science and artificial intelligence is applied to the problem of software representation to support the process of software change. A model of software to support change, suggested by the current research, is presented. Populating that model requires data from tasks related to code-based functional decomposition, data abstraction, and functional data flows. A competence model for the task of code-based functional decomposition is proposed.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"16 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":"131553692","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}
To improve the exchange of ideas, formalisms, and techniques among distributed artificial intelligence researchers, the authors have developed MICE as a standard and flexible testbed for experimentation. MICE allows an experimenter to specify the constraints and characteristics of an environment in which agents are simulated to act and interact, and does not assume any particular implementation of an agent's reasoning architecture. MICE therefore provides a platform for investigating and evaluating alternative reasoning architectures and coordination mechanisms in many different simulated environments. The authors outline the design of MICE and illustrate its flexibility by describing simulations that model coordination among heterogeneous agents in cooperative and competitive domains.<>
{"title":"Using MICE to study intelligent dynamic coordination","authors":"T. A. Montgomery, E. Durfee","doi":"10.1109/TAI.1990.130377","DOIUrl":"https://doi.org/10.1109/TAI.1990.130377","url":null,"abstract":"To improve the exchange of ideas, formalisms, and techniques among distributed artificial intelligence researchers, the authors have developed MICE as a standard and flexible testbed for experimentation. MICE allows an experimenter to specify the constraints and characteristics of an environment in which agents are simulated to act and interact, and does not assume any particular implementation of an agent's reasoning architecture. MICE therefore provides a platform for investigating and evaluating alternative reasoning architectures and coordination mechanisms in many different simulated environments. The authors outline the design of MICE and illustrate its flexibility by describing simulations that model coordination among heterogeneous agents in cooperative and competitive domains.<<ETX>>","PeriodicalId":366276,"journal":{"name":"[1990] Proceedings of the 2nd International IEEE Conference on Tools for Artificial Intelligence","volume":"119 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":"133423819","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：