Pattern orthogonalization via channel decorrelation by adaptive networks

The early processing of sensory information by neuronal circuits often includes a reshaping of activity patterns that may facilitate the further processing of stimulus representations in the brain. Motivated by recent studies in the olfactory system we study simple adaptive networks that aim to orthogonalize activity patterns representing similar stimuli. Biologically it is plausible that the adaptation is driven by simultaneous correlations between the input channels rather than by the similarity of input patterns that the animal experiences at different times. We demonstrate that networks can achieve effective pattern orthogonalization through channel decorrelation if they also equalize their output levels. In feedforward networks adaptation fails for even moderately similar input patterns. Recurrent networks do not have that limitation. When optimized for linear neural dynamics, they can orthogonalize the representations of highly similar input patterns quite well, even if a thresh...

• Type: Journal
• Year: 2010
• Where: JCNS
• Authors: Stuart D. Wick, Martin T. Wiechert, Rainer W. Friedrich, Hermann Riecke
• Keywords: Input Patterns, JCNS 2010, Neural Dynamics, Similar Input Patterns