Ray pavloski

Transcription

Ray pavloski
ARCHITETTURA AD ALGHERO
CONFERENZA – EVENTO DI FACOLTÀ
15 giugno 2010, ore 10,00
Asilo Sella, AIIPc, secondo piano aula grande
Lungomare Garibaldi, Alghero
Raymond Pavloski
Does the Form of the Synaptic Architecture
Yield Dimensions of Visual Perception?
Department of Psychology, Indiana University of Pennsylvania, USA
In spite of its increasing prominence as the focus of both empirical and theoretical investigations, the coexistence of private perceptual experience and familiar
objective measures of neural processes has no generally accepted explanation. It
is argued that three properties of a perceptual gestalt might be employed in
bridging the gap between these two domains: perceptual gestalts are hidden
from objective observation, they are stable, and they are organized at multiple
levels. Evidence from my recent research shows how simulations of model neural
networks produce self-organized patterns of clusters of neurons that are both
stable and hidden, and illustrates how the structure of these hidden patterns can
be inferred from the network-wide structure of the effects of source clusters on
target clusters. A new, categorical model of hidden patterns and its application to
simulation data will be presented. The maps and objects of this model describe
network-wide collections of neural processes and variables, respectively. It is predicted and confirmed by simulations that the form of the synaptic architecture
organizes action potential histories into stable, coarser-scale hidden patterns of
cluster states. In the model, the new cluster states emerge as a map representing
synaptic processes is factored into a pair of change-of-scale maps and a hidden
pattern object. The hidden pattern object is interpreted as the realization of an
information state consisting of cluster states that are chosen by network input
from an information space of possibilities defined by the network-wide form of
the synaptic architecture. In order to determine if the structure of this information state can be made isomorphic to the information content of a perceptual
gestalt, the information content of perceptual gestalts must be measured and
the complexity of the model must be increased.
Raymond Pavloski received the Ph.D. degree in experimental psychology from McMaster University, Canada. After completing postdoctoral research and supervised clinical
work supported by the Ontario Mental Health Foundation and the Ontario Heart Association, he joined the Special Professional Staff of the Department of Medicine, where
he conducted research in psychophysiology and served in the Behavioral Medicine
Unit. He is professor and chair of the Department of Psychology at Indiana University
of Pennsylvania, U.S.A. which he joined in 1984. His first efforts in neural network modeling led to the development of a self-trapping associative memory model. For the
past several years, his research has focused on investigating the relationship between
perceptual experience and activities in dynamic, richly-interconnected recurrent neural
networks. He was the recipient of an IUP Academic Excellence and Innovation Award
in support of his present research, which has the goal of developing a formal model
that can describe both the organization of perceptual gestalts and the organization of
hidden biological patterns produced by recurrent neural networks.