Stimulus Reconstruction from Nonlinear Neuronal Encoding
Year: 2002
Type: preprint
Abstract: The amount of information processed by the sensory system limits the extent of an organism’s perception. In consequence, stimulus reconstruction algorithms are of practical importance in quantifying the content of encoded sensory signals. Recent efforts in this way have made use of reverse Wiener (linear) filters to estimate the encoded stimulus via observation of the neural response. Experimental evidence, however, points to the existence of various non-linearities in the input-output behavior of the sensory pathway; it is therefore expected that non-linear estimation methods extract greater amounts of stimulus-related information from observed responses. Additionally, reverse filters do not provide information with regards to the forward encoding mechanism. Motivated as such, we present a two-stage framework to: (1) model the non-linear forward encoding process through a Volterra/Wiener series expansion, and (2) on this basis develop a Bayesian estimation algorithm for decoding. This approach, on the other hand, is directly linked with the forward encoding mechanisms and also improves upon the reverse-filter method in a variety of cases. Results also suggest that the above framework may be applicable in describing general non-linearities in neuronal encoding that have been observed experimentally in a variety of sensory pathways.
Source: SSRN Electronic Journal
Authors: Alireza Boloori, Garrett B. Stanley
Institutions: Michigan State University, Harvard University
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W180994371