INTEGRATION OF PROBLEM-SOLVING TECHNIQUES

Still another form of integration is the combining of two or more problem-solving techniques such as expert systems, fuzzy logic, neural nets, model-based reasoning, qualitative physics, simulators, case-based reasoning, and parallel computing. Most of these combinations are still in the prototyping and experimental stages. Nevertheless, they are indicative of the depth to which the technology is being explored in Japan.

Specific examples include:

  1. Fuzzy logic and neural networks: AdIn, a small Japanese ES company, has developed a prototype system that uses fuzzy membership functions within a neural network framework as a means of accelerating the learning process and improving problem solving performance. The application is the interpretation of images.
  2. Deep and shallow reasoning: a two-level expert system built by Toshiba (Chapter 2) is intended to diagnose unforeseen failures in a physical system. This system uses a shallow knowledge diagnostic system to handle foreseen failures, backed up by a deep knowledge system to tackle unforeseen problems. The deep knowledge system employs model-based reasoning and also employs a fuzzy simulator.
  3. Model-based reasoning and parallel computing: NEC showed us a prototype model-based diagnostic system operating on a parallel processor.
  4. Case-based and rule-based reasoning: a system developed at Toshiba which uses a recursive application of case-based reasoning; case-pieces are used to repair the mismatches in the primary case.
  5. CAD and knowledge acquisition: Toshiba has developed a knowledge acquisition tool in which a design process for highly stylized objects is acquired by monitoring a CAD system.

Published: May 1993; WTEC Hyper-Librarian