Tim@cis.upenn.edu (Tim Finin) (09/29/86)
CONNECTIONIST NETWORKS
Jerome A. Feldman
Computer Science Department
University of Rochester
There is a growing interest in highly interconnected networks of very simple
processing elements within artificial intelligence circles. These networks are
referred to as Connectionist Networks and are playing an increasingly important
role in artificial intelligence and cognitive science. This talk briefly
discusses the motivation behind pursuing the the connectionist approach, and
discusses a connectionist model of how mammals are able to deal with visual
objects and environments. The problems addressed include perceptual
constancies, eye movements and the stable visual world, object descriptions,
perceptual generalizations, and the representation of extrapersonal space.
The development is based on an action-oriented notion of perception. The
observer is assumed to be continuously sampling the ambient light for
information of current value. The central problem of vision is taken to be
categorizing and locating objects in the environment. The critical step in
this process is the linking of visual information to symbolic object
descriptions, i.e., indexing. The treatment focuses on the different
representations of information used in the visual system. The model employs
four representation frames that capture information in the following forms:
retinotopic, head-based, symbolic, and allocentric.
The talk ends with a discussion of how connectionist models are being realized
on existing architectures such as large multiprocessors.
Thursday, October 2, 1986
Room 216 - Moore School
3:00 - 4:30 p.m.
Refreshments Available
Faculty Lounge - 2:00 - 3:00 p.m.