-
Yunosuke HAGA
Article type: Preface
1996 Volume 11 Issue 4 Pages
497
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hiroshi YASUHARA
Article type: Cover article
1996 Volume 11 Issue 4 Pages
498-499
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Makoto NAGAO
Article type: Special issue
1996 Volume 11 Issue 4 Pages
500-506
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hozumi TANAKA
Article type: Special issue
1996 Volume 11 Issue 4 Pages
507-513
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hirosato NOMURA
Article type: Special issue
1996 Volume 11 Issue 4 Pages
514-521
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Kazunori MURAKI
Article type: Special issue
1996 Volume 11 Issue 4 Pages
522-529
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Jun-ichi TSUJII
Article type: Special issue
1996 Volume 11 Issue 4 Pages
530-541
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Masayuki NUMAO, Hiroshi MASUDA
Article type: Corner article
1996 Volume 11 Issue 4 Pages
542-549
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hiroaki KITANO
Article type: Corner article
1996 Volume 11 Issue 4 Pages
550-553
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hideyuki NAKASHIMA
Article type: Corner article
1996 Volume 11 Issue 4 Pages
553-555
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Hiroshi MOTODA
Article type: Corner article
1996 Volume 11 Issue 4 Pages
555-557
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Setsuo OHSUGA
Article type: Corner article
1996 Volume 11 Issue 4 Pages
557-559
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Osamu DAIRIKI
Article type: Corner article
1996 Volume 11 Issue 4 Pages
559-561
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Riichiro MIZOGUCHI
Article type: Corner article
1996 Volume 11 Issue 4 Pages
561-565
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Kiichi AKABA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
566-573
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
We propose a Symbolic Manipulation type Neural-network Model (SMNM), which organizes knowledge representation with association interconnecting symbols such as semantic network, and carry out knowledge retrieval by neural networks operation. That is a new type neural networks model consisting of localized and explicit representation of knowledge, and performing intelligence as symbolic manipulater. The model has same good points of its organization, that it is ability to manipulate a lot of knowledge and to reason with semantics and flexible execution. Also the representation with network of the model can be replaced with sets of symbols on the basis of its organization, and carry out knowledge retrieval by sets operation instead of by neural networks operation. That become a new type knowledge representation with sets and will participate activity by application to programming tool for AI. This paper describes about the organization of the model, the knowledge representation by sets, and application examples with the knowledge representation.
View full abstract
-
Takashi ONODA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
574-584
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
In engineering fields, one of the most important application of artificial neural networks is a non-linear parametric model that can approximate any continuous input-output relation. The quality of approximation depends on the architecture of the neural network used, as well as on the complexity of the target relation. Usually, we do not have accurate input-output relation and we can utilize only input-output examples. In such a case, if a fixed architecture of networks is given, the problem of finding a suitable set of parameters that approximate an unknown relation is usually solved using supervised learning algorithms. Supervised learning is carried out based on a training set which consists of a number of examples observed from an actual system. An important but difficult problem is to determine the optimal number of parameters. This paper presents a statistical approach to this problem of model selection, or determination of the number of hidden units. In other words, we describe a measure of neural networks reliability statistically. The relation between the training error, which is reduced by training process, and the generalization error, which is a measure of the quality of approximation, is derived by the following two gaps: ・One is a statistical gap between the optimal and the semi-optimal solution caused by the finite training examples. ・The other is a gap between the optimal and estimated solution caused by the learning process. In this paper, we describe that the gap between the optimal and the semi-optimal solution is measured by generalizing Akaike's information criterion (AIC) to be applicable to unfaithful models and the gap between the semi-optimal and estimated solution is estimated by a cross validation. The above gaps can explain the relation between the training error and the generalization error. This relation leads to a Neural Network Information Criterion (NNIC) which is useful for selecting the optimal network model based on a given training set. We present that this criterion is useful with a simple simulation.
View full abstract
-
Masaki KUTEMATSU, Takahira YAMAGUCHI
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
585-592
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
Although Case-Based Reasoning comes up in order to solve knowledge acquisition bottleneck, a case structure acquisition bottleneck emerges there in CBR instead of it. Because we cannot decide an appropriate case structure in advance, a framework for CBR should be able to improve a case structure dynamically, collecting and analyzing cases. Here is discussed a new framework for knowledge acquisition using CBR and model inference. Model Inference tries to obtain new descriptors (predicates) with interaction of a domain expert, regarding the predicates as the slots that compose a case structure, focusing on the function of predicate invention. The framework has two features: (1) CBR obtains a more suitable group of slots (a case structure) incrementally through cooperation with model inference, and (2) model inference with predicate invention capability discovers the rules which deal with a given task better. The system has been applied to the legal analogy problem to acquire new legal interpretation rules from given precedents. The system has invented two important legal predicates and generated two legal interpretation rules including some legal doctrine related to the problem. And the case structure has been improved using the two invented predicates. The experimental results show us that the framework is promising to acquire knowledge in the field of legal interpretation.
View full abstract
-
Takahira YAMAGUCHI, Masaki KUREMATSU, Naotake SHIMOZU, Hiroshi NAKAO, ...
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
593-599
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
We present a new framework for knowledge acquisition using CBR and model inference. Model Inference tries to obtain new descriptors (predicates) with interaction of a domain expert, regarding the predicates as the slots that compose a case structure, focusing on the function of predicate invention. The framework has two features: (1) CBR obtains a more suitable group of slots (a case structure)incrementally through cooperation with model inference, and (2) model inference with predicate invention capability discovers the rules which deal with a given task better. The system has been applied to SA/SD software development method. The system has invented a new predicate that is a so important to describe structural diagrams and generated three useful rules to modify them. And the case structure has been improved using the invented predicate. The experimental results show us that the framework is promising to acquire knowledge in the field of SA/SD software development method. Furthermore, we discuss future works about the framework.
View full abstract
-
Makoto TABUCHI, Toshiharu TAURA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
600-607
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
In this paper, a new method for knowledge acquisition is proposed. This method involves an interactive dialog between a learning engine and a human expert. Knowledge is hard for experts to recognize by themselves. They can, however, point out errors and recognize a lack of knowledge. Research in this field to date tends to improve reasoning or learning to obtain more precise knowledge. This approach, however, requires completeness of sample data and domain knowledge, which is hard to achieve. It is the opinion of the authors that the process should place a greater emphasis on the role of the experts. The method proposed in this paper is to sophisticate knowledge via the cycle of conversation between a deductive learning engine and a human. In this method, Genetic Programming (GP) is used as a deductive learning engine which provides great flexibility for interaction. GP shows a human the resulting learnt knowledge, of which he/she previously had only vague understanding before seeing. Next, he/she operates a fitness function of GP through choosing knowledge which he/she regards as good, to reflect them process of learning. In other words, it can be said that conditions are determined dynamically in the interaction between system and human. The authors conducted an experiment to prove the merit of this method using a prototype system.
View full abstract
-
Yoshiki KISHI, Satoshi KOIDO, Kazuhito SHIBATA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
608-618
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
T] and V, in which H, T and V mean Prolog variables. In bits blocks, each variable is treated as an address pointer and unification among variables is pointer addressing from a variable to another. The addressing is encoded to a bits block together with a delimiter at the places correspondent to the variable. The arguments of a goal are recognized in sevral continuous bits blocks, and arguments by bits are translated to Prolog terms according to meaning of each bit. Learning from the examples is executed by GA operations on a population of a given size and by a meta interpreter. The GA operations in use are the multiplex crossover whose points are decided at random, the uniform mutation which partially uses the adaptive mutation, and the elitist strategy. The meta interpreter examines an individual on possibility of translation to a Prolog clause and scores the ratio of satisfied examples against the givens as the fitness of the translated clause. If the individual of the highest fitness gives true to some examples, a clause has learned and the examples given true are deleted. The learning process runs adding one goal to the older clause step by step. If no example remains, learning ends. Applying the method to the 'append' learned its definition within about 30 generations of 20 elite individuals each.
View full abstract
-
Shigeyoshi TSUTSUI, Yoshiji FUJIMOTO
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
619-628
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
We propose a new type of multi-population GA, that is the p-fGA (Phenotypic Forking GA), an extension of the previously proposed g-fGA (Genotype Forking GA). We use multi-population schemes ; this includes one parent population with a blocking mode and one or more child populations with a shrinking mode. In the g-fGA, sub-space for each population is defined by the salient schema within the genotypic search space. In contrast to this, the p-fGA defines its sub-space by the neighborhood hypercube around the current best individual in the phenotypic search space. The empirical results show that the p-fGA has a good performance as well as the g-fGA, and the variable resolution p-fGA has a capability of searching with high resolution and can improve the local searching capability in a genetic search.
View full abstract
-
Shin"ichiro NISHIZAWA, Hiroshi NAKAGAWA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
629-636
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
One of the most important problem for natural language processing in Japanese is the zero anaphora resolution. We focus here on Japanese complex sentences conjoined by "node". According to [Nakagawa 94, Nakagawa 95], we can use a new pragmatic role called motivated and a traditional concept called "Point of View" to identify the coreference relations among semantic roles of subordinate and those of main clauses. These constrain the relation among semantic and pragmatic roles within subordinate or main clause for Japanese complex sentences. Then, we construct the system that treats this type of relation by constraint logic programming, and we use feature structures to formalize these constraints in unification grammar formalism. Finally, the result of an example sentence by the proposed system is shown.
View full abstract
-
Takehiko ABE, Haruhiko KIMURA, Hidetaka NANBO
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
637-644
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
Production systems are an established method for encoding knowledge in an expert system. A disadvantage of production systems is that they require a large amount of computation to perform the matching of the left-hand sides since they must not only determine whether or not a rule is satisfied, but also enumerate all ordered subsets of working memory elements satisfying a rule's left-hand side among all rules. In this paper we propose a match algorithm for production systems, which is a method for dealing with expensive productions in production systems. Expensive productions are rules which would be required the extraordinary time and space to match. This paper presents a method for reducing computer loads which are caused by expensive productions. To be more concrete, in order to decrease the number of combinations in matching, we propose a new data structure of α-memory such that, as a rule is given, it can immediately point out working memory elements satisfying the rule's condition element. Using this new α-memory, we present a method for dealing with expensive productions in production systems.
View full abstract
-
Seiji YAMADA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
645-652
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
This paper describes a novel method SIP to interleave planning with execution in a dynamic environment. To determine the timing of interleaving them, we uses the success probability, SP, that it successfully executes a plan in an environment. SP is formally defined with likelihood that all operators in a plan are executable in an environment, and we develop a method to compute it inexpensively. An interleave planning system integrates reactivity with deliberation depending on dynamics of an environment. We require planning for intelligent behavior, and need to integrate reactivity with deliberation. Unfortunately, few solution have been proposed to this problem. Our approach gives a solution by interleaving planning with execution. We assign input probabilities to effects of actions and persistence of objects in an environment. Plans are transformed into Bayesian networks on which their SPs are computed in O(n) time: n is plan size. A system switches planning to execution when SP falls below an execution threshold. After the execution, a system observes an environment, and starts planning again.
View full abstract
-
Kanji UCHINO, Hitoshi KANOH, Seiichi NISHIHARA
Article type: Technical paper
1996 Volume 11 Issue 4 Pages
653-661
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
We propose a new approach to understanding engineering drawings, which is the problem to recover the 3D data of the legal solid objects from the 2D data of a given three-view orthographic drawing. Our method essentially consists of two major processes: translation of the original 2D data to a constraint satisfaction problem (CSP) by using a constraint knowledge base, and solving the CSP to get the final 3D data by using a general CSP solver. To realize the translation process, we develop geometrical rules derived from the restoration knowledge based on the axiom of existence of 3D objects in the usual Euclidean space. Since a CSP, an intermediate expression of an original drawing, is actually a knowledge base independent from both the given 2D data and the final 3D data, our approach enables us (1) to tune up the knowledge base easily, (2) to customize the restoration system with high modularity, and (3) to make use of existing general purpose constraint solvers adopting many kinds of efficient search techniques as well.
View full abstract
-
[in Japanese]
Article type: Other
1996 Volume 11 Issue 4 Pages
662
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
[in Japanese]
Article type: Corner article
1996 Volume 11 Issue 4 Pages
663-665
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
[in Japanese]
Article type: Corner article
1996 Volume 11 Issue 4 Pages
665-666
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
[in Japanese]
Article type: Corner article
1996 Volume 11 Issue 4 Pages
667
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
[in Japanese]
Article type: Corner article
1996 Volume 11 Issue 4 Pages
668
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Activity report
1996 Volume 11 Issue 4 Pages
669-671
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Activity report
1996 Volume 11 Issue 4 Pages
672-674
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Activity report
1996 Volume 11 Issue 4 Pages
675-676
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Activity report
1996 Volume 11 Issue 4 Pages
b001-b016
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Cover page
1996 Volume 11 Issue 4 Pages
c004
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Cover page
1996 Volume 11 Issue 4 Pages
c004_2
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Table of contents
1996 Volume 11 Issue 4 Pages
i004
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS
-
Article type: Table of contents
1996 Volume 11 Issue 4 Pages
i004_2
Published: July 01, 1996
Released on J-STAGE: September 29, 2020
MAGAZINE
FREE ACCESS