Hirohisa Seki

Image filtering, edge detection, and edge tracing using fuzzy reasoning

We characterize the problem of detecting edges in images as a fuzzy reasoning problem. The edge detection problem is divided into three stages: filtering, detection, and tracing. Images are filtered by applying fuzzy reasoning based on local pixel characteristics to control the degree of Gaussian smoothing. Filtered images are then subjected to a simple edge detection algorithm which evaluates the edge fuzzy membership value for each pixel, based on local image characteristics. Finally, pixels having high edge membership are traced and assembled into structures, again using fuzzy reasoning to guide the tracing process. The filtering, detection, and tracing algorithms are tested on several test images. Comparison is made with a standard edge detection technique.

Verification of Prolog programs using an extension of execution

An approach to proving properties of Prolog programs exploiting characteristics of Prolog is described. The most important feature of this approach is the use of an extension of execution, which is a generalization of the conventional Prolog interpreter. We use the extended execution to show that a property S in a class of first order formulas, called S-formulas, is a logical consequence of the completion of a program P. This approach is (1) simple because we need only an extention of the Prolog interpreter, (2) understandable because properties are processed keeping their original forms as far as possible and (3) without waste because we carry it out without unnecessary explicit strengthening of P. We show how the extended execution works for the same example in the Boyer and Moore Theorem Prover (BMTP).

Partial Deduction of Disjunctive Logic Programs: A Declarative Approach

This paper presents a partial deduction method for disjunctive logic programs. We first show that standard partial deduction in logic programming is not applicable as it is in the context of disjunctive logic programs. Then we introduce a new partial deduction technique for disjunctive logic programs, and show that it preserves the minimal model semantics of positive disjunctive programs, and the stable model semantics of normal disjunctive programs. Goal-oriented partial deduction is also presented for query optimization.

Cost-based abduction using binary decision diagrams

This paper proposes an abductive reasoning system, which can find most preferable solution efficiently, using Binary Decision Diagrams. We propose a specialized BDD and its operation suitable for abductive reasoning: PBDD (Partial BDD) and GPC (Graft & Pruning Construction). We have implemented PBDD and GPC algorithm and built a cost-based abductive reasoning system which can find much more efficiently the most preferable explanation of a given observation. We have also made some experiments on the system with some diagnostic problems. Some good performance results are also shown.

Tow-down Induction of Logic Programs from Incomplete Samples

We propose an ILP system FOIL-I, which induces logic programs by a top-down method from incomplete samples. An incomplete sample is constituted by some of positive examples and negative examples on a finite domain. FOIL-I has an evaluation function to estimate candidate definitions, the function which is composition of an information-based function and an encoding complexity measure. FOILI uses a best-first search using the evaluation function to make use of suspicious but necessary candidates. Other particular points include a treatment for recursive definitions and removal of redundant clauses. Randomly selected incomplete samples are tested with FOIL-I, QuinIans FOIL and Muggletons Progol. Compared with others FOIL-I can induce target relations in many cases from small incomplete samples.

Parallel Cost-based Abductive Reasoning for Distributed Memory Systems

This paper describes efficient parallel first-order cost-based abductive reasoning for distributed memory systems. A search control technique of parallel best-first search is introduced into abductive reasoning mechanism, thereby finding much more efficiently a minimal-cost explanation of a given observation. We propose a PARallel Cost-based Abductive Reasoning system, PARCAR, and give an informal analysis of PARCAR. We also implement PARCAR on an MIMD distributed memory parallel computer, Fujitsu AP1000, and show some performance results.

On rule checking and learning in an acupuncture diagnosis fuzzy expert system by genetic algorithm

ADFL is an acupuncture diagnosis expert system based on fuzzy logic, which has been reported by Zhang et al. (1993, 1994). This paper gives a method to learn ADFL rules by genetic algorithms (GAs), and a technique to optimize and check fuzzy membership functions. For any length gene, we present an extended ordinal representation method. A fitness function described with fuzzy rules is explained. The experimental result indicates this method is effective.<<ETX>>

On inductive and coinductive proofs via unfold/fold transformations

We consider a new application condition of negative unfolding, which guarantees its safe use in unfold/fold transformation of stratified logic programs. The new condition of negative unfolding is a natural one, since it is considered as a special case of replacement rule. The correctness of our unfold/fold transformation system in the sense of the perfect model semantics is proved. We then consider the coinductive proof rules proposed by Jaffar et al. We show that our unfold/fold transformation system, when used together with Lloyd-Topor transformation, can prove a proof problem which is provable by the coinductive proof rules by Jaffar et al. To this end, we propose a new replacement rule, called sound replacement, which is not necessarily equivalence-preserving, but is essential to perform a reasoning step corresponding to coinduction.

A -SATCHMORE: SATCHMORE with availability checking

We present an improvement of SATCHMORE, calledA-SATCHMORE, by incorporating availability checking into relevancy. Because some atoms unavailable to the further computation are also marked relevant, SATCHMORE suffers from a potential explosion of the search space. Addressing this weakness of SATCHMORE, we show that an atom does not need to be marked relevant unless it is available to the further computation and no non-Horn clause needs to be selected unless all its consequent atoms are marked availably relevant, i.e., unless it is totally availably relevant. In this way,A-SATCHMORE is able to further restrict the ues of non-Horn clauses (therefore to reduce the search space) and makes the proof more goal-oriented. Our theorem prover,A-SATCHMORE, can be simply implemented in PROLOG based on SATCHMORE. We discuss how to incorporate availability cheeking into relevancy, describe our improvement and present the implementation. We also prove that our theorem prover is sound and complete, and provide examples to show the power of our availability approach.

A knitting pattern recognition and stitch symbol generating system for knit designing

Abstract In this paper, a system for supporting a knit designing is described. This system generates stitch symbols from a knitting pattern image. There are 24 kinds of stitch symbols. Each of them represents a basic simple pattern of a knitting. To implement this system on the computer, a method for recognizing string structure from a knitting pattern image and a method for representing string structures of the knitting pattern uniquely, are proposed. The utility of this system is verified from some execution examples.