著者

沼尾 正行 志村 正道

出版者

一般社団法人情報処理学会

雑誌

情報処理学会論文誌 (ISSN:18827764)

巻号頁・発行日

vol.26, no.2, pp.247-256, 1985-03-15

---

分散処理システムで グラフリタクション方式により関数型言語を評価するためには システムを構成する各プロセッサにグラフを分散して配置する必要がある.グラフが各プロセッサに分散していると リタクションを行うために他のプロセッサ内の記憶装置を参照したり 書き換えたりすることが必要となる.また 複数のプロセッサで一つのグラフを書き換えるため グラフのアクセスに対して危険領域を設定しなければならない.本論文では これらの煩雑な問題を解消するため ノード単位でリダクションを行う方法を述べる.この方法では プロセスがグラフの各ノードに割り当てられており アークを通して互いに通信し合う.これらのプロセスにより リタクションがノード単位で行われるので グラフを分割して各プロセッサに割り当てることが容易である.リタクションの基本操作はノードの消去とノードのコピーであり これらの二つの操作を組み合わせることにより リダクションが行われる.ノード単位でリダクションを行うことにより グラフを各プロセッサの共用データとする必要がなくなるため 共用データにアドレスを割りふったり アクセスを行うプログラムに危険領域を設けたりする必要がなくなる.このため 大規模で拡張性の高い分散リタクションシステムを構築することが可能となる.

著者

大谷 紀子 志村 正道

出版者

一般社団法人 人工知能学会

雑誌

人工知能学会論文誌 (ISSN:13460714)

巻号頁・発行日

vol.19, no.5, pp.399-404, 2004 (Released:2004-07-06)

参考文献数

11

被引用文献数

2 1

---

In representing classification rules by decision trees, simplicity of tree structure is as important as predictive accuracy especially in consideration of the comprehensibility to a human, the memory capacity and the time required to classify. Trees tend to be complex when they get high accuracy. This paper proposes a novel method for generating accurate and simple decision trees based on symbiotic evolution. It is distinctive of symbiotic evolution that two different populations are evolved in parallel through genetic algorithms. In our method one's individuals are partial trees of height 1, and the other's individuals are whole trees represented by the combinations of the former individuals. Generally, overfitting to training examples prevents getting high predictive accuracy. In order to circumvent this difficulty, individuals are evaluated with not only the accuracy in training examples but also the correct answer biased rate indicating the dispersion of the correct answers in the terminal nodes. Based on our method we developed a system called SESAT for generating decision trees. Our experimental results show that SESAT compares favorably with other systems on several datasets in the UCI repository. SESAT has the ability to generate more simple trees than C5.0 without sacrificing predictive accuracy.

著者

第4代会長 志村 正道

雑誌

人工知能

巻号頁・発行日

vol.31, no.4, 2016-07-01

著者

志村 正道

出版者

日本認知科学会

雑誌

認知科学 (ISSN:13417924)

巻号頁・発行日

vol.1, no.1, pp.1_3-1_4, 1994-05-20 (Released:2008-10-03)

著者

ブンサーム キッスィリクン 沼尾 正行 志村 正道

出版者

社団法人人工知能学会

雑誌

人工知能学会誌 (ISSN:09128085)

巻号頁・発行日

vol.8, no.1, pp.46-54, 1993-01-01

被引用文献数

2

---

Recently there has been an increasing interest in learning systems which induce first-order logic programs from examples. However, due to the intractability of the hypothesis space, such systems which use exhaustive search, e.g., MIS, are unable to learn reasonably complex programs. Previous solutions have been proposed to overcome this difficulty : while some systems restrict their hypothesis space, others use heuristics or additional knowledge such as analogical structures or abstraction. However, existing systems still have limitations. GOLEM gives up completeness by restricting the hypothesis space to only determinate clauses. A determinate clause is a clause composed of only determinate literals. A literal is determinate if each new variable in it has only one binding. Even the commonly used generate-and-test programs generate candidate solutions for their test routine, and thus are non-determinate. FOIL avoids exhaustive search by using Gain heuristic to select a literal that greedily discriminates positive examples from negative examples. Although it is able to learn some classes of problems efficiently, the heuristic fails to capture other aspects of usefulness of a literal, i.e., it overlooks a useful literal which produces no discrimination. For instance, a literalpartition (Head, Tail, List 1, List 2) in the quick-sort program does not discriminate between positive and negative examples. We propose a new heuristic-based approach to the learning of Horn-clause logic program with list structure. Our system, CHAM, learns a class of complex programs not learned by previous systems, i.e., non-determinate programs out of the learning space of GOLEM, and programs with non-discriminating literals which pose difficulties for FOIL. Experiments on learning Prolog programs with list structure have shown that while being able to learn a large class of programs, CHAM preserves efficiency in various test problems.

著者

ブンサーン キッスィリクン 沼尾 正行 志村 正道

出版者

社団法人人工知能学会

雑誌

人工知能学会誌 (ISSN:09128085)

巻号頁・発行日

vol.7, no.6, pp.1027-1037, 1992-11-01

被引用文献数

8

---

Learning systems find a concept based on positive and negative examples by using given terms, such as features and predicates. Most learning systems employ selective induction, and find a concept description composed of only predefined terms. However, if such terms are not appropriate, constructive induction or shift of bias are required to invent new terms. Although there has been increasing interest in systems which induce a first-order logic program from examples, there are few systems that perform constructive induction. FOCL invents new terms, i.e., new predicates, by combining existing subpredicates. Based on interaction with its user, CIGOL invents terms without any given subpredicate. This paper presents Discrimination-Based Constructive inductive learning (DBC) which invents a new predicate without any given subpredicate nor any user interaction. Triggered by failure of selective induction, DBC introduces a new predicate into a previously found incomplete clause. This is performed by searching for a minimal relevant variable set forming a new predicate that discriminates positive examples from negative ones. If necessary, DBC also recursively invents subpredicates for the definition. Experimental results show that, without interactive guidance, our system CHAMP can construct meaningful predicates on predefined ones or from scratch. Our approach is system independent and applicable to other selective learning systems such as FOIL.