{"id":346,"date":"2012-06-07T16:56:00","date_gmt":"2012-06-07T14:56:00","guid":{"rendered":"http:\/\/project.inria.fr\/keops\/?page_id=346"},"modified":"2012-07-24T17:17:24","modified_gmt":"2012-07-24T15:17:24","slug":"recherche-modeliser","status":"publish","type":"page","link":"https:\/\/project.inria.fr\/keops\/recherche-2\/recherche-modeliser\/","title":{"rendered":"Research : Modelling"},"content":{"rendered":"

Identifying non-linear mapping from natural images to non-standard sensor behavior.<\/span><\/strong><\/p>\n

Objective:<\/strong> Design and develop new functional models of non-linear local visual operators based on\u00a0sparse representations and independent component analysis methods in order to encounter for\u00a0sophisticated dynamical and statistical pre-processing modules of natural image sequences.
\nThis includes conductance-based modeling of motion direction selectivity observed in some retinal ganglion cells.
\nMethods:<\/strong> In order to better understand to which extends non-standard visual sensor responses\u00a0are able to process the visual signal using still non-elucidated mechanisms, we propose to explore\u00a0original non-linear local mapping from natural images using a variational approach at the\u00a0mesoscopic level.
\nThe general framework, already well-established and validated on non-trivial visual operators\u00a0(such as motion perception and segmentation, visual event detection, etc) allows one to build a\u00a0link between (i) high-level specification of how the brain represents and categorizes the causes of\u00a0its sensory input and (ii) related analog or spiking neural networks. Focusing on visual processing,\u00a0this computer-vision expertise allows one to show -for a rather general class of computations- how\u00a0it is possible to directly rely \u201cwhat is to be done” (perceptual task) with \u201chow to do it” (neural\u00a0network calculation). More precisely, in computer vision, efficient computation using\u00a0implementations of regularization processes allow one to obtain well-defined and powerful\u00a0estimations. They (i) represent what is to be done as an optimization problem, (ii) considering\u00a0regularization mechanisms (implemented using so-called partial-differential-equations) and (iii)\u00a0\u201ccompiling” the related analog or spiking neural network parameters. An unbiased approximation\u00a0of a so-called diffusion operator used in regularization mechanisms with a direct link between\u00a0continuous formulation and the related sampled implementation is available, and spiking-mechanisms have been explored in this context (Vieville et al, 2007).
\nThis includes non-linear local visual operators based on sparse representations. We propose to apply this general framework in the present context\u00a0in order to somehow reverse engineer the non-standard cells processing. The original framework\u00a0has to be revisited in order to design well-founded mechanism to learn the proper parameters,
\ngiven a set of input\/output. Multi-model estimation methods are going to be used, to guarantee the\u00a0estimation with a minimal number of parameters. Spike trains coding of information is going to be\u00a0considered here (Cessac et al, 2009).
\nThis modeling is not going to be a precise description of the internal retinal processes, but of its\u00a0input\/output relation. The key point is that we are going to be able to consider natural images\u00a0sequences (and not artificial biased stimuli) as input, and thus need more sophisticated algorithms\u00a0than for simple artificial stimuli.<\/p>\n

\"\"<\/a><\/p>\n

\"\"
\nTask steps:<\/strong>
\n(i) Formalization of the standard LN retinal transducer model as a variational process, and\u00a0analysis of the limits of this standard approach.
\n(ii) Generalization of the original method to sparse representations, high-level image statistics\u00a0operators (using, e.g. ICA).
\n(iii) Production of a specification for the Task 4 simulator.
\n(iv) Experimental spike-trains benchmarks, using data preprocessed by Task 3.<\/p>\n


\n<\/strong><\/p>","protected":false},"excerpt":{"rendered":"

Identifying non-linear mapping from natural images to non-standard sensor behavior. Objective: Design and develop new functional models of non-linear local visual operators based on\u00a0sparse representations and independent component analysis methods in order to encounter for\u00a0sophisticated dynamical and statistical pre-processing modules of natural image sequences. This includes conductance-based modeling of motion direction selectivity observed in some … <\/p>\n

Continue reading<\/a><\/p>\n","protected":false},"author":36,"featured_media":0,"parent":128,"menu_order":2,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-346","page","type-page","status-publish","hentry","nodate","item-wrap"],"_links":{"self":[{"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/pages\/346","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/users\/36"}],"replies":[{"embeddable":true,"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/comments?post=346"}],"version-history":[{"count":8,"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/pages\/346\/revisions"}],"predecessor-version":[{"id":649,"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/pages\/346\/revisions\/649"}],"up":[{"embeddable":true,"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/pages\/128"}],"wp:attachment":[{"href":"https:\/\/project.inria.fr\/keops\/wp-json\/wp\/v2\/media?parent=346"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}