Skip to Main content Skip to Navigation

Natural computation, much ado about nothing

Abstract : The information coding of the environment by the brain is subject to intense debate at the level of its biological implementation (functional properties) as well as at the level of its reliability, efficiency and precision. This study is based on in-vivo intracellular recordings of responses of primary visual cortex neurons to stimuli of various complexity including natural scene animated by eye-movements. We show that functional selectivity are generated by cortical recurrence notably by the excitation/inhibition balance expressed through a large diversity of combinatory of synaptic input functional selectivity. In natural condition with eye-movement, the observed code is highly reliable, binary and temporal and cortical computation is removing input redundancies. Nonlinearity or gain controls, which biphasic temporal profile sculpts the cortical responses, increase the temporal precision and reproducibility of the code. We propose in discussion, a sketch of a statistical model of adaptation and self-organisation consisting in the maximisation of the mutual-information between the system and its environment. This model, based on a decomposition of entropy into microscopic ordered redundancy, is sufficient to account for some defining properties of living systems such as reliability, diversity/specialisation, and optimal or efficient coding.
Document type :
Complete list of metadatas
Contributor : Pierre Baudot <>
Submitted on : Friday, January 11, 2008 - 9:40:17 AM
Last modification on : Wednesday, December 9, 2020 - 3:11:51 PM
Long-term archiving on: : Tuesday, April 13, 2010 - 5:01:35 PM


  • HAL Id : tel-00203712, version 1


Pierre Baudot. Natural computation, much ado about nothing. Neurons and Cognition [q-bio.NC]. Université Pierre et Marie Curie - Paris VI, 2006. English. ⟨tel-00203712⟩



Record views


Files downloads