Author:
Raymond Pavloski
Affiliation:
Indiana Univ. of PA, United States
Keyword(s):
Hard Problem, Neurotechnology, Recurrent Neural Network, Sentience, Tolerance Space, Visual Object Unity.
Related
Ontology
Subjects/Areas/Topics:
Health Engineering and Technology Applications
;
Neurocomputing
;
Neurotechnology, Electronics and Informatics
Abstract:
Achieving an understanding of how qualities of experience arise in concert with the operation of neural networks could produce a revolutionary advance in neurotechnology. The work reported here explores a relationship between a visual quality and neural activity that has not previously been investigated: visual object unity may emerge from and constrain neural interactions. Simulations were employed to determine if a topological signature of a unified object develops as a recurrent neural network’s activity is modulated by retinal input. Results show that differences in recurrent excitatory conductance values produced by adjacent active neurons are negligibly small, and can be described by a tolerance relation. Tolerance open balls about the vectors of conductance values produced by active neurons emerge in response to the retinal image of an object and a connected open set consisting of intersecting open balls quickly develops. Such connected open sets are invariant over fluctuation
s in participating neurons, demonstrate several characteristics of perception, and are hypothesized to be objective signatures of perceived object unity. Dynamical network phenomena, such as hysteresis, lead to empirical predictions that can be tested with human participants. Means of identifying objective signatures in brain activity are considered.
(More)