Direction des Relations Internationales (DRI)
NOTA : La proposition d'Equipe Associée comporte du Chilien plusieurs partenaires fédérés autour de l'Université de Valparaiso pour simplifier la gestion de cette EA.
La proposition en bref / The proposal in brief
Much progress has been made in the last decades in understanding the basic organization and function of the nervous system in general. Contributions to this end have come from various domains including computational neuroscience and numerical science of the information in general.
1. Objectifs scientifiques de la proposition / Scientific goals of the proposal (1 to 2 pages)
Description of the proposal's objectives and brief perspective regarding the state-of-the-art:
We propose to apply computational methods to model the behavior (spike response) of retinal ganglion cells in response to natural images. This last knowledge will use to further develop our artificial retina model created at the INRIA. The results will be incorporate into real engineering applications, as new dynamical early-visual modules, for artificial vision devices.
The retina, an accessible part of the brain, is a unique model for studying how neurons code visual signals from natural scenarios. A recent review proposes that visual functions (e.g. movement, orientation, anticipatory temporal prediction, contrast), thought to be the exclusive duty of higher brain centers, are actually carried at the retina level. The anatomical and physiological segregation of visual scenes into spatial, temporal and chromatic channels begins at the retina through the action of local neural networks. However, how the precise articulation of this neural network contributes to numerical local solution at the bases of global perception remains in general a mystery.
We propose to model the complexity of behaviors find in retinal ganglion cells (the output to the brain) in front of natural images and incorporate the results into our artificial retina model. Typically, the “classical” behavior of retinal ganglion cells has been evaluated using artificial stimuli as drifting gratings or flashes. But, ganglion cells produce unexpected behavior when the input stimulus is replaced by natural images (or movies), which is what vertebrate’s eye daily see and represent the real natural early processing capacity of the eye. We revisit both the retinal neural coding information sent to the brain, and at the same time, the development of new engineering applications inspired by the understanding of such neural encoding mechanisms. We develop an innovative formalism that takes the real (natural) complexity of retinal responses into account. We also develop new dynamical early-visual modules necessary to solve visual problems task.
Short description of the scientific tasks planned for the three years:
2. Présentation des partenaires / Partners presentation
UV-CINV: The Universidad de Valparaiso team groups members of the Faculty of Science from the area of Neuroscience, biomedicine and biostatistics. They have been working together and some of their members collaborate already with INRIA team’s fellows. The UV-CNV team is interested in the study of the sensory neural coding process and its formal computational models as well as their implementation in robotics and artificial vision devices. We work in basic research using sensory biology (Palacios and Orio) from single to multineuronal recording in the nervous system; computational neuroscience and statistics (Guiraud, Orio, Salas); EEG and PET modelling (Chabert) and Robotic (Zagal). The UV team has several ongoing project: 1) Recording from single cells and multielectrodes in a large population of retinal neurons with the finality to study principles for neural coding 2) Developing a Robotic approach using concepts of embodiment and sensory constraint; 3) Modeling the locomotors trajectory of a rodent model for degenerative diseases (e.g. Alzheimer) were we are exploring the use of entropy and where a Bayesian model to characterize the degree of memory loses; 4) Models of diffusion magnetic resonance imaging to understand the origin of diffusion signal to providea assistance in the interpretation of diffusion MR images. The groups met regularly and have several students in co-tutelle.
Palacios Vargas 51 years old,
nationality, Professor Universidad de Valparaiso, 2008-2010
Visiting Researcher CORTEX INRIA and CREA Ecole Polytechnique, France
Education Ph.D. in Neurosciences (1991) Université de Pierre et Marie Paris VI, France; 1990-1997 Postdoctoral and Associate Research Scientist, Yale U, USA.
Research domains: Sensory systems must ultimately be understood at different biological levels, from molecular events to animal behavior seen in their natural ecological conditions. With Francisco Varela we show that Avian (Columba livia) uses ultraviolet light to shape its color vision space (Vision Research, 32:1947-1953, 1992) and we propose a theoretical framework “Ways of coloring” to this aims published in Behavioral and Brain Sciences, 15:1-74, 1992 (along 33 commentaries). At Yale I was working to establish the spectral sensitivities and kinetics response of rods and cones by recording photocurrents with suction electrodes (J Physiol (London). 471:817-829. 1993). We reported UV cones in Danio aequipinnatus, a small cyprinid related to zebrafish (Vis Neuroscien. 13:411-421. 1996, Vis Res. 38:2135-2146. 1998). Moving to the Universidad de Valparaiso in 1997 I started my own lab, focussing in visual sensory ecology from single photoreceptor, electroretinogram (ERG) and reflectance characterization from animals and their microhabitat, that are likely to be used as signals in sexual selection, recognition of conspecifics, camouflage. In retinas slices using patch-clamp, we are studying the retina neural network form by bipolar, amacrines and ganglions cells types and their participation in color vision. More recently we are recording from a population of ganglion cells using multielectrodes MEA array (64X). Furthermore, at the retinal an intricate bioelectrical neural network develop dynamically depending on the background illumination and the behavioral meaning of the task for the animal. We are using mathematical tools to model retinal neural coding. Recently, we start to work in the field of Neurobiology of Learning and Memory, using behavior, biochemistry and synaptic plasticity (LTP, LTD) approaches in Octodon degus, a rodent that during aging develop brain marks proper of Alzheimer diseases. Another area of my interest is complexity and I coordinating several multidisciplinary activities in the area of cognitive sciences at the Instituto de Sistema de Complejos de Valparaiso, Chile (www.iscv.cl). For more details see http://www.cnv.cl/palacios
Peichl L, Chavez AE, Ocampo A, Mena W, Bozinovic F, Palacios AG (2005). Eye and Vision in the Subterranean Rodent Cururo (Spalacopus Cyanus, Octodontidae). J Comp Neurol 486:197-208.
Goles E, Palacios AG (2007) Dynamical complexity in cognitive neural networks. Biol Res 40:381-384
Delgado LM; Vielma AH; Palacios AG; Schmachtenberg O. (2009) The GABAergic system in the retina of neonate and adult Octodon degus, studied by immunohistochemistry and electroretinography J Comp Neurol. 514(5):459-472.
Schleich C, Vielma A, Palacios AG, Peichl L. (2010) The Retinal Photoreceptors of Two Subterranean Tuco-tuco Species (Rodentia, Ctenomys): Morphology, Topography and Spectral Sensitivity. In press J Comp Neurology.
Palacios AG, Bozinovic F, Vielma A, Arrese CA, David M. Hunt DM, Peichl L (2010) Retinal Photoreceptor Arrangement, SWS1 and LWS Opsin Sequence, and Electroretinography in the South American Marsupial Thylamys elegans (Waterhouse, 1839). J Comp Neurology 518(9):1589-1602.
Author of 49 paper, participation in several divulgation activities, 101 meeting abstract, supervision 5 Postdoctoral fellows; thesis 12 graduate and 10 undergraduate. Coordinator of institutional external grants (MECESUP, Fundación Andes). Member of study sections from FONDECYT and Students Fellowships during 6 years. Member of the steering committee MECESUP2 (from Chilean Education Minister, 2006-2010). Coordinator Research grants 3 Fondecyt, 3 DIPUV, 1 NIH-FIRCA, Co-director Milenio Nucleus, coordinator PBCT ring ACT45 Coordinator of a French ECOS-CONICYT project and co-investigator in a European .project. MORPHEX. Co-investigator CNRS Neuroinformatic (2009-2010) (www.cnv.cl/palacios).
goal of the research of the CORTEX team is to study the properties
and capacities of distributed, numerical and adaptive automated
information processing and to show that this kind of processing may
allow to build "intelligent" systems, i.e. able to extract
knowledge from data and to manipulate that knowledge to solve
problems. More precisely, these studies rely on the elaboration and
analysis of neuromimetic connectionist models, developed along two
sources of inspiration, computational neuroscience and machine
Working in computational neuroscience, the following objectives have been considered:
1. At the interface between the microscopic and mesoscopic levels, precise and realistic models of neurons and of their related network dynamics allow us to analyze the neural code in Spiking Neurons and Networks and to better understand the role of such neuronal characteristics as the timing of spikes and spike's assemblies, including the interplay between inhibitory and excitatory activities.
2. At the interface between mesoscopic and macroscopic scales, we model populations of neurons using diferent models of Neural Fields, in order to better understand the functioning and learning characteristics of a local circuit of neurons, seen as a high level unit of computation. This level of description allows us to study the interaction between such calculation maps, i.e., large neuronal systems, such as cerebral maps and their feedbacks, as observed in sensorimotor loops.
3. In order to remain consistent with biological and ecological characteristics, we develop embodied and Embedded Systems: From a behavioral point of view, the emerging cognition has to be situated, i.e. resulting from a real interaction in the long term with the environment.
From a computational point of view, computations have to be really distributed, with decentralized control and memory.
In order to be falsiable through the confrontation to the previous constraints, our models have to be embodied in systems (e.g. robots) that interact with their multimodal environment, and embedded in parallel architectures of computations (eg FPGA, clusters).
4. Equally important, constraints relative to the consistency with experimental data have to be considered. We devote our activity of database analysis and interpretation, on physiological signals and psychometric data recorded from the functioning brain (eg EEG, ECOG). This common work with experimental neuroscience brings us a more macroscopic view of cerebral regions implicated in Higher Level Functions and allows us to design Brain-Machine Interfaces (BMI).
In association with CORTEX, the
NEUROMATHCOMP team, regarding this calloboration, aims at stuying
motion perpection models (in biological and artificial vision,
considering spiking neural networks and variational and PDE-based
approaches) and methods combining dynamical
systems theory, statistical physics and ergodic theory allowing to
classify dynamics arising in canonical neuronal networks models like
integrate and fire models, including synaptic and intrinsic
plasticity, spike coding, spike train statistics analysis, neural
masses dynamics, with applications in vision and imaging.
Furthermore, CORTEX and NEUROMATHCOMP have join
their efforts to develop common software tools in these fields, which
aregoing to be of primary use in this project.
years old, French, Director of Research 2nd class INRIA, French
1997: habilitation to be a director of research, university Henri Poincaré, Nancy1: "Intelligence neuromimétique"
1990: PhD in computer science, university Henri Poincaré, Nancy1 : "Une modélisation fonctionnelle du cortex: la colonne corticale"
1986: Engineer diploma, INPL
2000-present: DR INRIA
1990-2000: CR INRIA
Involvement in the research community
Member (and moderator) of the Scientific Committee of the inititative NeuroComp (french community in Computational Neuroscience)
Member of the steering committee of the PIRSTEC ANR program (survey and prospective analysis in cognitive science in France)
Responsible of CogniEst, french East Network in Cognitive Science
Head of the EPI CORTEX (Computational Neuroscience) at INRIA
Kassab, R., Alexandre. F. (2009) Incremental Data-driven Learning of a Novelty Detection Model for One-Class Classification with Application to High-Dimensional Noisy Data, Machine Learning, 74(2), 191-234.
J. Vitay, N.P. Rougier and F. Alexandre, 2005. A distributed model of visual spatial attention. in: Biomimetic Neural Learning for Intelligent Robotics. S. Wermter, G. Palm and M. Elshaw Eds. Springer, 54-72.
J.C. Sarrazin, A. Tonnelier, F. Alexandre, 2005, A model of contextual effect on reproduced extents in recall tasks: the issue of the imputed motion hypothesis, Biological Cybernetics, 92(5), 303-315.
F. Alexandre, 2009, Cortical basis of communication: local computation, coordination, attention. Neural Networks, 22 (2) 126-133.
J. Fix, N. Rougier, F. Alexandre, From physiological principles to computational models of the cortex, Journal of Physiology, 101, 1-3, pp. 32-39, 2007.
Topics of research: Computational Neurosciences, cortex modeling, cerebral models of architecture and learning in the visual system, distributed computation, machine learning, artificial intelligence, memory, learning and reasoning, cognitive science
Supervision of 25 PhD students, responsible of 1 European project, 1 ANR project, 20 articles, 10 chapters, 60 international conferences.
List, for each partner, the researchers involved in the Associate Team, with a short resume of the Coordinator;
Frederic Alexandre-INRIA is a Research Senior at INRIA. He is the head of the CORTEX Research Team at INRIA Nancy Research Center. His domain of research is Computational Neuroscience: through the study of cerebral information flows, neuronal architecture and learning principles, his main question is to understand how complex intelligent behavior emerges from distributed neuronal computation. He supervised more than 25 PhD students and was implicated in several international projects, including the leadership of a European project. He is also implicated in a variety of French committees, including the NeuroComp initiative about Computational Neuroscience and the PIRSTEC survey about Cognitive Sciences and Technologies in France.
Bruno Cessac-INRIA PhD in Physic. He work as a theoretical physicist on modeling and analysis of large sized dynamical systems and especially neuronal networks dynamics. He has developed methods combining dynamical systems theory, statistical physics and ergodic theory allowing characterizing the dynamics arising in canonical neuronal networks at the microscopic (neuron dynamics) and mesoscopic level (neural masses). His expertise is crucial, in this project, regarding the use of sophisticated statistical methods. The cornerstone of the analysis performed here is his method, coming from ergodic theory, and already available at the implementation level.
Thierry Viéville-INRIA is a Researcher Senior at INRIA where he works in Computational Neurosciences, while he teaches and advices PhD students at the Nice Sophia-Antipolis University. His research interests after Computer Vision is now Computation Neurosciences, more precisely Visual Perception, Motion Analysis and Adaptive Processes. He advised more than ten PhD students and participated in several international collaborations (EEC projects) with WP responsibilities. He also helps the INRIA board regarding Scientific Culture.
Pierre Kornprobst -INRIA obtained his PhD in Mathematics 1998. Then joined the computer science department from University of Southern California (Los Angeles), working with Gérard Medioni as a CSNE (Coopérant au Service National en Entreprise) sponsored by the company MATRA Système et Information. Since 2000, he is a researcher at INRIA Sophia Antipolis, participating to several project teams and defended his Habilitation in 2007. His research interests are computational and biological vision, computational neuroscience, psychophysics, calculus of variations, nonlinear partial differential equations and numerical analysis as applied to image processing. He is the key person, in this project, regarding mathematical aspects of variational approaches, and links with other connected research projects
Adrian Palacios-UV-CINV (AP-CINV, (Universidad de Valparaiso (UV) Professor since 1996) PhD in Neuroscience Pierre et Marie curie Paris VI, 1990 Fellowship Foundations: Philippe and Simone & Cino del Duca, France; 1990-1997 Postdoctoral and Associate Research Scientist, Yale U. 1998 Fellow Fundacion Andes; Visiting Research-Professor: Mind/Brain Institute, Johns Hopkins, 2001 MCB Harvard U. Fellows 2003 Center for Advanced Studies in Ecology & Biodiversity (CASEB); 2008-2010 Professor and Research Visiting INRIA-Cortex and Fellow to CREA Ecole Polytecnique, France. Founder and former director of a Master and a PhD Neuroscience Programs UV. 2006 Graduate coordinator, Faculty of Science UV. With Eric Goles and others they created in 2002 the Institute for Complex Systems of Valparaiso (ISCV) Author of 45 paper, participation in several divulgation activities, 102 meeting abstract, supervision 5 Postdoctoral fellows; thesis 12 graduate and 10 undergraduate. Coordinator of several public institutional external grants (MECESUP, Fundación Andes). Member of study sections from FONDEYCT and Students Fellowships during 6 years. Actual member Steering committee MECESUP2 (from Chilean Education Minister). Coordinator of several Research grants 3 Fondecyt, 3 DIPUV, 1 NIH-FIRCA, Co-director Milenio Nucleus, Principal investigador and coordinator PBCT ring grants ACT45 Coordinator of a French ECOS-CONICYT project and co-investigator in a European project MORPHEX and CNRS-Neuroinformatic (2009-2011) proyect. (www.cnv.cl, www.iscv.cl).
Laurent Bougrain -INRIA is an associate professor at Nancy-university working on neural information processing. He is working on a top-down approach for which data analysis techniques extract properties of underlying neural activity to better understand the principles of network dynamics. He studies the spatial and temporal scales of EEG, LFP and spikes in particular, for brain-machine interfaces. He is the scientific referent for health and disability at INRIA. He is the winner of the brain-computer interface competition IV, dataset 4 on "finger flexion from ECoG". He is the international coordinator of a STIC-AmSud projet (2009-2010), Argentine et Chili (UV) on single-trial detection for brain-computer interface.
Nicolas Rougier-INRIA: is an experienced researcher at INRIA and is a member of the CORTEX Research Team at INRIA Nancy Research Center. His domain of research relates to computational neuroscience with a focus on distributed numerical adaptive computations in order to understand how cognition can emerge from such computations. He has recently supervised two PhD thesis on the computational modeling on visual attention. He has been involved in various national, European and international projects.
Axel Hutt-INRIA is a theoretical physicist working on the synchronization of measured neural activity data and their modeling by neural population models. His work deals with continuous neuronal networks, which are extended in the spatial domain. The theoretical studies investigate and the spatio-temporal activity and networks with respect to effects of propagation delay, feedback delay and random fluctuations (noise). The investigated models consider several cell types, excitatory and inhibitory synapses and various types of spatial connectivities. He leads the CNRS NeuroInformatique project 2009-2010: "sensory transduction to perception".
Steren Chabert UV
Universidad de Valparaíso (UV). Professor since 2005, in the
Biomedical Engineering Department. Founder and director of a Master
program in Biomedical Engineering, at UV (since 2009). Director
of a FONDECYT research grant, focused on diffusion MRI, of a DIPUV
research grant (UV internal grants), collaborator in a CORFO Innova
project, in a DIPUV, and in an international collaboration project
STIC-AMsud. Author of 8 papers, 27 meeting abstracts. 8 undergraduate
thesis directed. Education: 2004: Ph.D. in Biomedical Engineering
from Université de Technologie de Compiegne, thesis directed by Dr.
D. Lebihan in CEA, on diffusion MRI. 2004-2005: postdoctorate
fellowship from ECOS-Conycit, in Dr. P. Irarrázaval Laboratory on
diffusion MRI. 2000: Master of Science in Biomedical Engineering at
Washington University in St Louis, MO (USA).
Guiraud-UV is Assistant Professor
Depart Statistic, UV since 2006. PhD in Theoretical Physics
Universite de Provence 2004, Marseille France. 2004 -2006
Postdoctoral position in the Depart Ing Matematica, Univ Chile,
CMM Santiago de Chile. My principal line of research deals with high
dimensional dynamical systems and their applications in the modelling
of systems of interacting units. My PhD. thesis deals with Coupled
map lattices which are lattice discrete time dynamical systems, used
for instance in the modelling of chains particles, genes regulatory
networks and more generally reaction diffusion systems. In
Neuroscience I am working on the dynamics of neural networks,
especially integrate and fire neural networks.
Juan Cristobal Zagal-UCH
PhD is on Evolutionary Robotics and on going research is on
Self-Modeling with Hod Lipson at Cornell University. His research
deals with questions such as How to make robots to be more resilient?
How to speed up learning in robots? To answer these questions he has
been working in connection with Neuroscientists like Adrian Palacios
at UV and Per E. Roland at Karolinska Institute. The ideas of
Enactivism that are present in this project are of particular
interest for his research. Juan Cristobal has developed physics
simulators for various robots. A related publication won a best paper
award during the 2004 RoboCup international symposium. Juan Cristobal
has designed and implemented experiments with real robots as well as
the required simulation of them.
Patricio Orio-UV-CINV is Universidad de Valparaíso, professor since 2007). PhD in Molecular & Cell Biology & Neurosciences, Universidad de Chile 2004. 2005-2006 postdoctoral research at the Instituto de Neurociencias de Alicante, Spain. 2008 DIPUV (UV) Grant awarded, 2009 Fondecyt Grant awarded for initiation into research. Author or coauthor of 10 papers (4 first author) and 20 meeting abstracts. Teaching in undergraduate and PhD programs in the Universidad de Valparaíso. Current research interest: Mathematical analysis and modeling of neuronal excitability with emphasis in oscillatory phenomena and sensory encoding.
Rodrigo Salas UV: PhD current research is advocated to the development and theoretical analysis of the robustness and flexibility capabilities of the learning algorithms of artificial neural networks and neuro-fuzzy models. He is exploring biologically plausible algorithms to enhance the capabilities of current techniques. He also co-manage a Biomedical Engineering undergraduate program at the Valparaíso University, (12 professors and 250 students) thus allowing this consortium to disseminate his results also by teaching, in this structure and beyond.
For each partner, involvement of students in the proposal. Estimation of the number of students concerned and whether joint thesis supervision is expected ;
Carolina Saavedra, Horacio Rostro-Gonzales, Thomas Girod, Maxime Rio,
Wahiba Taouali, Computational Neuroscience PhD ongoing thesis work.
We plan to involve two others graduate students and Master students
under joint supervision.
UV-CINV and UFSM Claudio Elgueta, Carolina Soto, Erick Olivares, Neuroscience PhD ongoing thesis work. Luis Sanz, Informatics PhD ongoing thesis work at the Santa María University. Nicolás Moreno Statistics Master student from University of Valparaíso. Other studens Jaime Oliva (Master Eng. Biomed.), Herman Zepeda (Master Eng. Biomed.), Mariela Hidalgo ((Eng. Civil Biomed.). We will involve two others graduate students under joint supervision. Two other Enginnering students Joaquin Delgadillo from the Universidad Santa Maria in Valparaiso and Diego Pardo the Universidad de Valparaiso, Biomedical Engineering will apply for an internship to complete or start a Master degree at the INRIA in 2011.
Background of the collaboration between the teams ;
For several years, there is an ongoing
tradition, but somehow asymmetrical, of exchange between INRIA
research labs and Chilean labs. Many Chilean students, especially in
the area of Engineering, are doing (or have done) labs rotation, as
well as developing their thesis work in France. Recently, a very
dynamic program of internationalization of Chilean Science, with a
funding increase for international programs has helped to build more
balanced, human and resources exchange between both countries. Many
French Scientists and Students consider now Chile as a good place for
collaboration; some are actors in this project. Furthermore, since
2007, we have strengthened and developed several ongoing and past
collaborations and exchanges between INRIA and Chilean researcher’s
partners and students belonging to several Universities, in
particular from Valparaiso. An informal, but rapidly increasing set
of collaborative activities is emerging, targeting long-term team
associations. For example, Frédéric Alexandre and Thierry Viéville
taught in the 2008 ISCV-summer school in Valparaiso, and in the
ISCV-summer school in Valparaiso, with also Laurent Bougrain and Axel
Hutt, while Bruno Cessac is going to teach a course this autumn in
Valpapraiso, etc... Thanks to both INRIA and CMV-UV founds. Several
seminars (using video-conferences) between INRIA and labs in
Valparaiso have been issued. We now have a weekly
on our common subjects. In 2008, Dr. Palacios
participated as a external reviewer in Adrien Worher PhD Thesis
carried in the Odyssee team et INRIA. Then becomes an Associate
Visitor Professor for several months at INRIA-CORTEX and CREA-Ecole
Polytechnique. Two young French colleagues (Pierre Guiraud and Steren
Chabert at the UV), in link with this project, have permanent
positions in Valparaiso. More than 10 Chilean students have made an
Internship or PhD in the INRIA teams linked with this project.
Frédéric Alexandre, Laurent Bougrain, Steren Chaber and
Rodrigo Salas are involved in the two years STIC-AmSud project
2009-2010 "robust single-trial evoked potential detection for
brain-computer interfaces using computational intelligence
techniques", including one Internship and one Ph.D. thesis in
France. This geographic complementarity has also become a thematic
complementarity. A main impact of this project in Chile will be to
bring first class French research in mathematics, computation,
robotics and its application in benefices of students and researchers
from engineering, biology, experimental psychology and robotic. With
reverse transfer of competence towards France.
In summary, proposal combines efforts from researchers belonging to INRIA and two related French Universities with two Chilean Universities and one Research Institute, which represents a significant opportunity to build and develop a strong collaborative STIC platform for learning, sharing, transferring and applying knowledge in divers areas of S&T with a strong benefice for higher education and the society in general. In fact, thanks to previous founds the "CORTINA team" has already started ist action, while the goal is now to make this dynamic visible and sustainable.
Links towards relevant personal webpages, laboratories, home institutions, etc
3. Impact (1 page maximum) / Impact (maximum 1 page)
Impact of the proposed collaboration.
In a nutshell, this EA is going to bring to our every
day common work between Chili and France three add-on:
1/ Provide, on the French side, the mandatory ressources to maintain this collaboration's bundle (these links are supported for two years on Chilean resources).
2/ Make our work more formally visible and structured, in order to apply together to common financing on our research topics, including common publications.
3/ Allow partners to structure in the perspective of an INRIA Center of Excellence in Chili (Applied Computational Neuroscience will be a ligne to be incorporate (expected) during 2011-2012)
One outcome is going to be the organisation of common summer schools (after 2008 and 2010) with a higher visibility of the French participation.
The complementarity with respect to the other CorTexMex EA (leaded by Bernard Girau, Cortex and M.A. Arias-Estrada from Mexico) on hardware/software codesign of bio-inspired connectionist models for vision, is two-fold: the research corresponds to the dual topic of Cortex (bio-inspired systems versus computation neuroscience models) and to the work of colleagues not involved in CORTINA, except one, whose role is going to make the link between these two topics. Furthermore, despite the geographical complementarity (Mexico versus Chile), we would be pleased to embed these two collaborations in some more global coordinate actions (e.g. international conference) with ``Latin America´´, if this becomes pertinent in our field.
- the scientific objectives of participating teams;
In the area of STIC and its applications, like computational brain science and robotics are relatively scarce in Chile, in contrast to the existence of excellent engineering and computer science school. In France, INRIA and its partners, which aim at combining scientific excellence with technology transfer, can help. On the other hand, experimental brain and sensory sciences are better developed in Chilean labs, in the domains targeted here. This is the case for the retina physiology, where the CINV is in link with the worldwide best labs.
- the relationships between partners and their home institutions
The partners as a team in this proposal have many complementarities, from interdisciplinary fields that will facilitate the accomplishment of this application (i) a team (AP) from sensory and biophysics neuroscience team that will provide experimental data (single and multielectrodes cells recording) on retina physiology; (ii) experts (INRIA) on neural computation and population coding statistic to formalize how neural coding is created. Regarding experimental data, provided here as true entries for the interpretation and further modeling and robotic applications, we indeed will also consider some of the data, e.g. neuron biophysics properties, recollected from the literature or the present colleagues. And concentrate on the specific data set that cannot be borrowed from the literature, since obtained using new paradigms, with a strict control of the preparation and the natural stimuli, designed by us to answer particular question about temporal and spatial properties detailed previously.
On both sides, from this schematized context view, a win-win strategy emerges. In the area of STIC and its applications, like computational brain science and robotics, research teams are relatively scarce in Chile, in contrast to the existence of excellent engineering and computer science school. In France, INRIA and its partners, which aim at combining scientific excellence with technology transfer, can help. On the other hand, as mentionned before, experimental brain and sensory sciences are better developed in Chilean labs, in the domains targeted here. This is the case for the retina physiology, where ISCV and CNV are in link with the worldwide best labs. This is also the case for emerging topics in evolutionary robotics, where competences from several disciplines being aggregated.
4. Divers : néant.
Description du programme
scientifiquede travail (1 à 2 pages
/Description of the scientific work programme (maximum 1 to 2 pages)
Recent advances in multi-electrodes recording have thus brought us closer to understanding how populations of retinal ganglion cells encode visual information. By monitoring the visual responses of many ganglion cells at once, it is now possible to examine how ganglion cells act together to encode a visual scene. To attain this objective, a quantitative and statistical analysis of the ganglion cells spiking activity is required.
This issue is faced to the delicate problem of proposing and validating accurate statistical model fitting the empirical spike trains. It has been shown in (Schneidman et al, 2006; Cessac et al, 2008) that Gibbs measures constitute optimal parametric models, the estimated Gibbs potential allowing to produce population rate, correlations or synchronization pattern, providing an effective statistical tool.
Since, using the Gibbs potential framework allows us to obtain parametric estimations of spike train observables, e.g. the population rate, correlations, or synchronization pattern occurrence probability, this statistical tool appears to be a very interesting way of attaining our objective, allowing us to relate the observed spiking activity to higher scales of observation of the neuronal activity. For instance, the population spiking rate and correlation, or even higher order statistics can be measured using the previous parametric model and then integrated in mean-field mesoscopic models (Faugeras et al, 2009).
library (EnaS) which estimates a
polynomial Gibbs potential over population spike trains and
subsequently the population firing rate, correlations, higher order
statistics and relative entropy (Vasquez et al, 2010).
Two types of population spike trains are studied :
T3,1 : Simulated spike.trains of well-defined statistics in order to evaluate the quality and precision of the method, and,
T3.2 : Experimental spike-trains provided by CINV in order to evaluate the pertinence and the applicability of the method at the biological level.
already validated at the programmatic
level (functional tests). The goal here is to deal with the spike
decoding mechanism, more precisely to apply the proposed estimation
procedure of spike statistics in order to shed some light on the
correlations between neurons in the retina. This is a key question
because one hypothesis is that ganglion cells mainly act as independent
encoders (Segev et al, 2004). This would have important consequences
regarding computational aspects of the early-vision processing. This
hypothesis contradicts in part the results of another group (Pillow et
al, 2008), which has observed correlations between ganglion cells
encodings. As a perspective of this work, the teams are strongly
interested to attack the problem of retinal encoding of natural images.
The opportunity to combine experimental recordings and the best
numerical methods to analyze the data is a great advantage to attack
The Gantt Chart is strainghtforward since
all three tasks are going to be in constant interactions:
In words, T1 and T2 have
already started, while T2 is going to start as soon as data analysis is
going to available, while experimental work is planned for two years,
data analysis and modeling for two years and a half.
H., Vasquez J.C. and Viéville, T.
(2008), Statistics of spikes trains, synaptic plasticity and Gibbs
distributions, NeuroComp'08, Marseille.
Cessac, B., Paugam-Moisy H., Viéville, T. (2009), Overview of facts and issues about neural coding by spikes, J. Physiol Paris, 104, 1-2.
Escobar M.J., Masson G.S., Vieville T. and Kornprobst P. (2009). Action Recognition Using a Bio-Inspired Feedforward Spiking Network. International Journal of Computer Vision, 82-3 284-301.
Faugeras O, Touboul J, Cessac B (2009) A constructive mean field analysis of multi population neural networks with random synaptic weights and stochastic inputs” Front. Comput. Neurosci. 3:1.
Pillow W., Shlens J., Paninski L., Sher A., Litke A.M., Chichilnisky E.J. and Simoncelli E.P.(2008), Nature 454(7206), 995-999 .
Schwartz, G., and Berry, M.J., 2nd. (2008). Sophisticated temporal pattern recognition in retinal ganglion cells. J. Neurophysiol. 99, 1787–1798.
Schwartz, G., Harris, R., Shrom, D., and Berry, M.J., 2nd. (2007a). Detection and prediction of periodic patterns by the retina. Nat. Neurosci. 10, 552–554.
Schwartz, G., Taylor, S., Fisher, C., Harris, R., and Berry, M.J., 2nd. (2007b). Synchronized firing among retinal ganglion cells signals motion reversal. Neuron 55, 958–969.
Schneidman, E., Berry, M. J., Segev, R., and Bialek, W. (2006). Weak pairwise correlations imply strongly correlated network states in a neural population. Nature. 440, 1007—1012.
Segev R, Goodhouse J, Puchalla J, Berry MJ 2nd. (2004) Recording spikes from a large fraction of the ganglion cells in a retinal patch. Nat Neurosci. 7:1154-61.
Vasquez J,C, Viéville T, Cessac B (2010) Entropy-based parametric estimation of spike train statistics. J. Comp. Neuroscience (submitted).
Viéville, T., Chemla, S., and Kornprobst, P. (2007). How do high-level specifications of the brain relate to variational approaches? J Physiol Paris, 101(1-3):118-135.
1. Echanges / Exchanges
Incoming and outgoing exchanges planned: invitations of researchers from the partner institution in France, and missions of INRIA researchers abroad
Planned invitations and exchanges:
Status of the researchers involved (intern, PhD student, postdoctoral fellow, senior researcher, other)
Exchanges and join research and workshop are dedicated to both the participation to join international events (summer-schools, conferences, ..) and the works in the lab on experimental acquisition campain, where the physical presence is mandatory. All other common work is done using electronic tools, and does not require physical transportation.
Summary ofp the informations above and estimated budget
Note: Le calcul se base sur 3 éléments arithmétiques simples: voyage low-cost Chili-France 1.2K€ ; séjour (en tenant compte du fait que pour réduire les coûts et renforcer le travail déquipe, plusieurs collègues sont invités de manière familiale) 100€/jour x 20 jours ; surtout la plus grande partie de nos interactions se fait en visio-conférence et avec nous outils usuels de travail collaboratif numérique, ce qui limite fortement le rapport entre le coût de transport (et impact carbone) et la bande passante de cette collaboration.
2. Cofinancement / Cofinancing
1) CNRS NeuroInformatique project 2009-2010: "sensory transduction to perception" 20.000€. INRIA CORTEX: Frédéric Alexandre, Bruno Cessac, Laurent Bougrain, Axel Hutt (PI), Thierry Vieville; UV-CNV Chile: Adrian Palacios, Diego Cosmelli.
2) STIC AMSUD 09STIC01 project 2009-2010: "Robust single-trial evoked potential detection for brain-computer interfaces using computational intelligence techniques" 15.000€/2years. INRIA Nancy - France: Laurent Bougrain(PI), Frédéric Alexandre, Axel Hutt, Universidad de Valparaíso - Chile: Steren Chabert, Rodrigo Salas .
If this application is successful, partner institution will also support:
4) The Universidad de Valparaiso has yearly
application for Professor Exchange and Chilean graduate student can
apply to CONICYT for short periods of exchange with partners
5) Up to now, the Universidad de Valparaiso has supported all different travel costs related to summer school organization and scientific travels aboard.
3. Demande budgétaire / Proposed budget
© INRIA - mise à jour le 17/09/2010