CHIN%PLU@IO.ARC.NASA.GOV (03/23/88)
***************************************************************************
National Aeronautics and Space Administration
Ames Research Center
SEMINAR ANNOUNCEMENT
SPEAKER: Mr. Brian C. Falkenhainer
University of Illinois
TOPIC: Learning From Physical Analogies
ABSTRACT:
To make programs that understand and interact with the world as well as
people do, we must duplicate the kind of flexibility people exhibit when
conjecturing plausible explanations of the diverse physical phenomena they
encounter. We view this flexibility as arising from an ability to detect
similarities, within and across domains, between the various observed
behaviors. Interpreting an observation often requires the flexible integration
of knowledge from diverse sources and the formation of new theories about
the world. For example, understanding processes such as heat flow and
diffusion may involve reference to known theories of liquid flow, while
explaining the behavior of an oscillating LC electric circuit may require
a knowledge of springs.
Verification-Based Analogical Learning is an approach to theory formation
and revision which relies on analogical inference to hypothesize new theories,
and gedanken experiments (i.e., simulation) to analyze their validity.
It is an iterative process of hypothesis formation, verification, and revision
which focuses on the problem of validating analogically derived models.
This talk will describe the basic elements of verification-based analogical
learning, the kinds of flexible yet constrained reasoning they enable, and
discusses its implications for analogical reasoning in general. A number
of examples from the current implementation, PHINEAS, will be used to explain
and demonstrate the utility and diversity of this approach.
BIOGRAPHY:
Mr. Falkenhainer is a graduate student in the Ph.D program in Philosophy in
Computer Science at the University of Illinois, and is a Research
Assistant in the Qualitative Reasoning Group. His research in artificial
intelligence focuses on the general tasks of theory formation and observation
interpretation. A paper which appeared in the journal, Machine Learning,
summarized the results of his master's thesis on the discovery of functional
relationships in numeric data. A general tool for performing various types
of analogical mappings, called the Structure-Mapping Engine (SME), is
extensively described and analyzed in a paper recently submitted to the
journal, Artificial Intelligence. In support of SME, a probabilistic
generalization to traditional truth-maintenance systems was developed and
is described in a paper from the 1986 workshop on uncertainty in AI.
DATE: Monday, March 28, 1988 TIME: 3:00 - 4:00 pm BLDG. 244 Room 209
--------------
POINT OF CONTACT: Marlene Chin PHONE NUMBER: (415) 694-6525
NET ADDRESS: chin%plu@ames-io.arpa
***************************************************************************
VISITORS ARE WELCOME: Register and obtain vehicle pass at Ames Visitor
Reception Building (N-253) or the Security Station near Gate 18. Do not
use the Navy Main Gate.
Non-citizens (except Permanent Residents) must have prior approval from the
Director's Office one week in advance. Submit requests to the point of
contact indicated above. Non-citizens must register at the Visitor
Reception Building. Permanent Residents are required to show Alien
Registration Card at the time of registration.
***************************************************************************