Abstract
This paper shows how the distributed representation techniques
used in neural networks and other connectionist systems have a natural
counterpart in classifier systems. Eventually, this representational
correspondence may have important practical implications for the
parsimonious design of hybrid systems having both subsymbolic and symbolic
capabilities.
Abstract
As the technology for building knowledge based systems has
matured, important lessons have been learned about the relationship
between architecture of a system and the nature of the problems it is
intended to solve. We are implementing a knowledge engineering tool
called BaRT that is designed with these lessons in mind. BaRT is a
Bayesian reasoning tool that makes belief networks and other probabilistic
techniques available to knowledge engineers building classificatory
problem solvers. BaRT has already been used to develop a decision aid for
classifying ship images, and it is currently being used to manage
uncertainty in systems concerned with analyzing intelligence reports.
This paper discusses how state-of-the-art probabilistic methods fit
naturally into a knowledge based approach to classificatory problem
solving, and describes the current capabilities of BaRT.
Abstract
Many classification problems must be performed in a timely or time
constrained manner. For this reason, the generation of control schemes
which are capable of responding in real-time are fundamental to many
applications. For our problem, that is ship classification, tactical
scenarios often dictate the response time required from a system.
In this paper we discuss efficient ways to prioritize and gather
evidence within belief networks. We also suggest ways in which we can
structure our large problem into a series of small ones. This both
re-defines much of our control strategy into the system structure and also
localizes our run-time control issues into much smaller networks. The
overall control strategy thus includes the combination of both of these
methods. By combining them correctly we can reduce of the amount of
dynamic computation required during run-time, and thus improve thee
responsiveness of the system.
Abstract
In this paper a new mathematical procedure is presented for
combining different pieces of evidence which are represented in the
interval form to reflect our knowledge about the truth of a hypothesis.
Evidences may be correlated to each other (dependent evidences) or
conflicting in supports (conflicting evidences). First, assuming
independent evidences, we propose a methodology to construct combination
rules which obey a set of essential properties. The method is based on a
geometric model. We compare results obtained from Dempster-Shafer's rule
and the proposed combination rules with both conflicting and
non-conflicting data and show that the values generated by proposed
combining rules are in tune with our intuition in both cases. Secondly,
in the case that evidences are known to be dependent, we consider
extensions of the rules derived for handling conflicting evidence. The
performance of proposed rules are shown by different examples. The
results show that the proposed rules reasonably make decision under
dependent evidences.
Abstract
In this paper we present a new mathematical procedure for
combining conflicting evidence which are represented in the interval form.
We propose a methodology to construct combination rules which obey a set
of essential properties. he method is based on geometric model. We
compare results obtained from Dempster's, intervals Bayes and the proposed
combination rule with both conflicting and non-conflicting data and show
that the values generated by proposed combining rule are in tune with our
intuition.
Abstract
Testing and evaluation of an expert system are critical parts of
the expert system's life-cycle development. Often, the procedures and
results of these processes are not well documented. This paper presents
the preliminary testing and evaluation methods, observations, and results
for an expert system called CESA. It is hoped that the study will help
the knowledge engineer by serving as a documented example of testing and
evaluation procedures.
Abstract
FIS, short for Fault Isolation System, is an ongoing research
project at the Navy Center for Applied Research in Artificial
Intelligence, a branch of the US Naval Research Laboratory. The focus of
the work is model-based expert system shell capable of acquiring from a
user a description of a piece of electronic equipment called a unit under
test (UUT). This description is later used by FIS to perform such
diagnostic functions as recommending the next "best" test to make on the
UUT during a fault isolation sequence and estimating fault probabilities
after each test is made. These and other capabilities make FIS a powerful
tool which can be used in a variety of diagnostic settings, as discussed
in section 1.2. FIS has been developed primarily with large scale analog
hybrid electronic systems such as radar and sonar systems in mind, but is
applicable at least in principle to any human-engineered system with
discrete replaceable components.
FIS is written in LISP and has been under development for several
years. Detailed descriptions of the FIS system can be found in 1.2.3.4.
In this paper we provide an overview of the current system and describe
some of its current application areas.
Abstract
Evaluation of CALL systems depends on the answers to a series of
questions: What are our goals and priorities for language learning, and
within them what is demanded of CALL? What other kinds of entities -
video-tapes, human tutors, other software - do we implicitly or explicitly
set up as standards of comparison for CALL? To what extent do we want the
CALL system to fit with existing approaches and theories? Shall we
evaluate a CALL system as a monolith or by module? I elaborate these
questions with particular attention to intelligent tutoring systems, offer
some partial answers, and show how systems, learners, and theories are all
important objects of evaluation.
Abstract
Exploration can be an effective learning experience, if suitably
constrained and guided. Moreover, it can provide this benefit for
specifically targeted formal skills in the arithmetic curriculum. This
paper presents two complementary techniques for promoting success in
computer-based exploration environments. Semantic constraints on
exploration cut out meaningless options, and heuristic guidance
facilitates search on the basis of a heuristic function with both
cognitive and problem-solving components. We have implemented an
environment that permits semantically constrained exploration for
subtraction, as well as a related environment that facilitates the
transition to paper-and-pencil subtraction. The authors tested the system
in individual hour-long sessions with more than twenty children in grades
1-3.
Abstract
This paper describes the development of CESA, an expert system for
aiding in Defense research contracting. Contracting is a ripe area for
expert system application. Laypersons, such as scientists or new contract
specialists, typically have difficulties in understanding, synthesizing,
and applying relevant rules and regulations in the procurement request
generation process. CESA is designed to act as an advisory system for
aiding in the pre-award phase of the contractural process.
Abstract
Throughout U.S. Government agencies there are Contracting Officer
Technical Representatives (COTRs) who are in charge of monitoring
contracts and solving technical problems relating to these contracts.
COTRs are a diverse community ranging from physicists and chemists to
engineers and computer scientists. Automated support, and in particular
expert systems, might be one approach to aid the COTR in better executing
his/her duties and responsibilities. This article takes a look at using
expert systems to help the COTR, and also a selection methodology to
decide upom the best COTR problem domain for expert system development.
Abstract
An algorithm to perform belief maintenance was implemented on the
Butterfly Plus* Parallel Processor. This algorithm, which handles
reasoning with uncertainty, is used in a system called BaRT which performs
classification problem solving. The belief maintenance scheme uses a
network to represent the problem domain, where each node in the network
represents a hypothesis of the domain. The belief updating scheme is
inherently parallel; incoming evidence can be attached to any number of
different nodes in the network, and the impact of the evidence can be
propagated through the network in parallel. Results show that a
substantial improvement in the processing speed of belief updating can be
realized, especially in cases where a great del of evidence is entered
into the system at one time.
Abstract
The Naval Research Laboratory (NRL), Washington DC, began
operation in 1923 in response to a suggestion by Thomas Alva Edison that
"The government should maintain a great research laboratory...In this
could be developed...all the techniques of military and naval progression
without any great expense." NRL initially contained two research
division: radio and sound. It now consists of 15 research divisions
encompassing a very wide range of technical specialities. One of NRL's
newest research groups is the Navy Center for Applied Research in
Artificial Intelligence (NCARAI), where significant effort is being made
to transition artificial intelligence technology out of the laboratory and
into service. Many of the research projects conducted by NCARAI are
producing expert systems technology directed at broad problems within the
Navy. Several of these systems are now mature and the efforts made in
selecting problem domains, development, packaging, and distribution may
provide useful insights for others.
Intelligent M4 Systems
Machine Learning
Neural Networks
Sensor-Based Systems
Back to NCARAI Library