SENSOPAC
Structure de mise en forme 2 colonnes
  • Thursday 19 November 2015
  • table mise en forme contenu normal \xE0 gauche + boite de contenu \xE0 droite

    UPMC

    Contact:

    Dr. Angelo Arleo
    Laboratory of Neurobiology of Adaptive Processes
    Department of Life Science
    CNRS - University Pierre&Marie Curie
    box 14, 9 quai St. Bernard, 75005 Paris, France
    Phone: +33 1 44 27 32 54
    Mobile: +33 6 89 89 07 23
    Fax: +33 1 44 27 25 84
    email: angelo.arleo(at)snv.jussieu.fr

    Tasks:

    »  Neural encoding/decoding of hapic data
    »  Neural information processing/transfer at the granular layer of the cerebellum

    Progress

    A biologically-plausible encoding/decoding process accounting for the relative spike timing of the signals conveyed by the peripheral nerve fibres onto the second-order neurons in the cuneate nucleus (CN) has been studied. The rationale beneath this work is that the CN would not constitute a mere synaptic relay, but it would rather convey an optimal contextual account of tactile information to downstream modules (in particular the cerebellum) of the overall haptic system (see Module 5). Thus, the network developed during this work will play a relevant role in facilitating fast discrimination of haptic contexts, minimising destructive interference over lifelong learning, and maximising memory capacity, i.e. it will be relevant to the efficiency of the overall haptic classification process.A biologically-plausible encoding/decoding process accounting for the relative spike timing of the signals conveyed by the peripheral nerve fibres onto the second-order neurons in the cuneate nucleus (CN) has been studied. The rationale beneath this work is that the CN would not constitute a mere synaptic relay, but it would rather convey an optimal contextual account of tactile information to downstream modules (in particular the cerebellum) of the overall haptic system (see Module 5). Thus, the network developed during this work will play a relevant role in facilitating fast discrimination of haptic contexts, minimising destructive interference over lifelong learning, and maximising memory capacity, i.e. it will be relevant to the efficiency of the overall haptic classification process.

    A spiking neural network model has been developed and validated on a set of human microneurography data. The network model actively adapts its connectivity layout over time to maximize the information transfer at the CN level. A learning rule was implemented to this purpose, which yielded to efficient encoding/decoding performances (in terms of both fast and reliable haptic discrimination capabilities).

    An information-theoretic approach was set forth to assess the efficiency of the haptic classification system. Classical literature solutions based on Shannon\x92s mutual information (MI) do not permit to assess information transmission by taking into full account the metrics of the spike response space. A novel entropy definition was derived analytically that proved to be a promising tool to generalise the classical Shannon\x92s entropy for the encoding/decoding of spiking signals, and it allowed us to compute mutual information in the presence of a population of 450 spike trains with a 1 ms temporal precision.