On models of signal processing by neuronal patterns
Leang, Charles Chonglin
Johnson, Don H.
Master of Arts
A theory of information flow through sensory nervous systems is formulated as the representation of a finite number of events by a set of parallel point process channels with limited signaling sets. As a balance among traditional neural coding schemes that exploit neuronal specificity or the temporal structure of neural activity, a "channel state" coding scheme is proposed, in which the pattern of group neuron activities is defined as the empirical distribution of the channel states. From a detection approach, we propose several models and analyze them according to a system efficiency measure, the M-hypothesis Chernoff distance, which is shown to be the underlying vanishing rate of various estimation criteria. The models support the notion of flexibility, plasticity, survivability of sensory nervous systems, and can provide explanation for behavioral experience in signaling time, attention and perceptual precision.
Neurosciences; Electronics; Electrical engineering; Statistics