Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


Palermo handles this as a planning process like the I2SAdvisor system (Bremdal 95). Given a set of desired properties, it tries to construct a plan that produces a state with desired properties. In order to handle this, it applies a mixed planning paradigm:

  • Case-based reasoning (CBR)
  • Derivational analogy
  • Goal regression

The CBR part contains a case base of plans. Historic plans can be recognized and retrieved and applied as a prototype for the new problem. The derivational analogy part is a searched-based problem-solving unit that operates on a case base of derivations. Derivations suggest what planning operator to choose at a given step in the planning, if a similar result in the past resulted in the derivation of a successful plan. The goal regression part is constituted by a more search-based, problem-solving approach in the STRIPS tradition. It posts goals and works at them one at a time. When sub-goal interactions are discovered, it backtracks and reorders the sub-goals. Palermo handles a conjunction of goals. Each query is treated as a kind of goal conjunction. Each conjunct is attacked independently. The end results from each are combined. When the plan is constructed, a controller assigns tasks to specific processing agents and coordinates their execution. The agents are together responsible for achieving the goals that the plan pursues. SEIDAM and the agents communicate using sockets and the planning operators that send commands to the third-party software and await responses. The choice of operators is based on training observed when a human domain expert has performed similar interactions with the third-party software. For each step of the solution, the agents report on the state of the processing. These reports let SEIDAM learn about its successes and failures. In this manner it can refine its case base and its knowledge of the processing agents.

SEIDAM runs on a SUN Sparcstation using the Solaris 2.3 version of UNIX.

4. WILDLIFE TAXONOMY

4.1. INTRODUCTION

When Winston introduced his book on LISP, he also initiated a rule-based system on animal taxonomy (Winston 81). Although meant as an illustration on how to create a rule-based system in LISP, the accompanied example served as a fine depiction on how to do capture taxonomic knowledge in a rule-based model. A rule would typically read:

IF animal has pointed teeth
  AND animal has claws
  AND animal has forward eyes
  THEN animal is a carnivore

The simplicity of the approach inspired many people and students of expert systems at the time. Over the years we must assume that multiple prototypes of the kind were made, yet few seem to have become more than that. The initiatives that Jeffers (Jeffers 85) reports are, however, typical. Correct identification of species of plants and animals is necessary for ecological management. But the taxonomic systems with a professional inclination tend to be of value for purely biological purposes. Over the years neural networks have augmented the traditional rule-based approach. Neural networks applications tend to bring the support aspect much further. Instead of expecting structured information from an intelligent user, neural networks applications try to interpret much more unrefined data. In 1994, Boddy (Boddy 94) wrote about two systems that had successfully applied neural networks technology. One system was able to identify and classify 40 species of phytoplankton and fungal species from flow cytometry data. Another system was used for distinguishing between European and Africanized bees based on training on morphometric data. These type of applications, although addressing a complicated field, are pursuing the technology application aspect in much the same manner as the mineral analytic systems. Several varieties of neural networks have been tried with much the same conclusions as that for other types of applications. The training and vectorization problems are the hardest to resolve. Since the knowledge lies implicit in the network, their inherent knowledge structures are indistinguishable from those in geology that we have previously discussed.

We will report on an expert system that builds directly on the PROSPECTOR tradition. It is unique in its simplicity, yet has a very profound and powerful use potential. It is particularly relevant to the field we are discussing here, where large and complex systems dominate. The Whale Watcher Expert System should serve as an instance of the saying "keep it simple and beautiful." The Acquire expert system for whale identification is a system that addresses a problem that concerns many professionals within animal taxation and wildlife management. The approach that Acquire uses may well be as important to the field as the use of numbered identification rings and tags to mark species of birds, fish, and animals.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.