About

Being able to emulate the human brain (or approximations or parts of it) is one of the next major biological/technological breakthroughs that will have impacts in our daily life (health, autonomous machines, artificial intelligence, etc.). If recent progresses in machine learning have roots in bio-inspired techniques, mostly dated from the 70's, they usually represent very simple and specific structures without catching the essence of structure-function brain mechanisms. In machine learning, computer scientists build usually bio-inspired algorithms but not transdisciplinary biocognitive algorithms. In biology and cognition, researches are limited by the lack of experimental data (only lacunar data can be collected on animals and even less on humans). Our main goal is to develop a computational brain tool as an intermediate layer between high-level features (behavioral and cognitive functions at the human level) and low-level features (biological layers studied in mice, i.e., an animal brain that can be a model of human brain, as well as modifications observed in pathologies).

Objectives

To develop a computational brain tool as an intermediate layer between high-level features (behavioral and cognitive functions at the human level) and low-level features (biological layers studied in mice, i.e., an animal brain that can be a model of human brain, as well as modifications observed in pathologies).

To achieve this goal, we will focus on the hippocampus and the cortex, two brain structures highly implicated in learning. Here, learning is identified as the core integrative concept of the brain, but its complex processing within neuronal networks is still little understood. Using previous data obtained by biologists in this consortium and new biological data specifically obtained during this project, learning rules will be mathematized and implemented from synapses to behavioral and cognitive functions. In the long term, the computational brain can be conceived as a new local research tool providing in silico experiments to complete brain biocognitive data providing insights in neuronal learning modeling. In particular, this new tool will be used to study corresponding diseases (Alzheimer and epilepsy), to foster pedagogical innovation, and to develop new machine learning algorithms and artificial intelligence. 

 

More specifically, our common tasks are to: 

  • Dispose of a new in vivo biocomputational platform for cognitive/behavioral experiments to correlate the activity of local neuronal circuits (local field potential (LFP), single-unit recordings) with global measurements achieved by electroencephalography (EEG) and behavioral actions achieved at mouse/human levels.
  • Model human learning mechanisms based on specific neuronal circuits with biologically-realistic synaptic learning.
  • Achieve a new efficient reusable neurocognitive machine learning algorithm.
  • Dispose of a mathematical reusable model of neurocognitive machine learning and use it to extend current deep learning algorithms, e.g., testing their mathematical realization to electronic gates.

 

At the national level, the strengths of UCAJEDI concern mathematics, computer science, and biology. Fully integrated in this perspective, we propose ComputaBrain, federates the main local actors in the domains of neurobiology, mathematics, computer science, and cognition (IBV, IPMC, LJAD, I3S, INRIA and BCL) through this unique opportunity. This consortium constitutes a strong local concentration of modelers and simulationists collaborating via this trans-disciplinary project.

Workplan

The project is divided into 5 major work packages (WP). Consortium interactions between biocognitive experimental levels and mathematical/computational modeling is shown in Figure below.

ComputaBrain Project Structure

In both cognitive and neurobiological sciences, experiments are already performed in Nice Sophia Antipolis at different scales (from molecular mechanisms to individual behaviors) using different devices (whole-cell recording, Multi-Electrode Arrays (MEA), Local Field Potential (LFP), Electroencephalography (EEG), learning tasks, etc.). Most of these devices are used at animal scale (mouse), but some are also used at human scale (EEG, response times, error rates). Our goal is to extend current local biological experimental techniques to be able to fully and multimodally collect data from learning mice and to bridge the gap with human behavior and cognition through neural network modeling. Methodologically, the biocognitive experiments will be set in a microscale (synapse, neuron, circuits) to full organism approach. To increase pertinence of work, comparable experiments will be defined at behavioral cognitive level between mice and humans (cf. “Learning behaviors”). Corresponding circuits (focusing on hippocampus and cortex) will be identified to set experiments at this microscale level. Modelers will constantly interact with biologists and cognitivists to ensure the adequacy between the data to be provided and both analysis and modeling.

Partners' expertise

At IPMC, Hélène Marie's team is a long-term expert in mechanisms of synaptic and neuronal plasticity associated to learning using whole-cell patch clamping on the one hand (e.g., Marie et al., 2005; Middei et al., 2013) and spatial organism-level learning using rodent behavioral tasks (e.g., Tse et al., 2007; Bethus et al. 2010) on the other hand, but no experimental links currently exist within the team between these two levels (purpose of the new IPMC platform proposed here). Massimo Mantegazza's team is expert in neuron excitability investigations at the single cell level by whole-cell recordings and in studies of neuronal activity by EEG recordings in rodents (Heidrich et al., 2014; Cestele et al 2013; Liautard et al., 2013, 2014). Both Hélène Marie and Massimo Mantegazza teams will provide knowledge on biological mechanisms of brain circuits in healthy and disease conditions (Alzheimer, epilepsy) (eg. Marchetti et al. 2011; Guerrini et al. 2014). At iBV, Michel Studer's team will provide knowledge on cortical circuit development and assembly (Alfano et al., 2014; Harb et al., 2016) and is currently developing microcircuit analysis ex vivo using multi-electrode arrays (MEA) in collaboration with LJAD and I3S. At I3S, Alexandre Muzy's project recently extended (theoretically and experimentally) the theory of sequential machines using a new kind of algorithms correlating the activity of structure-based components with the whole network behavior (Muzy & Zeigler, 2014a, 2014b) but lacking theoretical proofs for complex network structures. Alexandre Muzy's project, Modélisation, Simulation & Neurocognition (MS&N) (I3S and LJAD), emerged from previous project Bio-info at I3S. At LJAD, in NeuroStatMod, researchers have expertise in sequential learning and optimization (Devaine et al., 2013), which allows comparing theoretically learning algorithms but lacks network structures. Besides, several statistical methods have been developed to analyze neuronal activity, notably for single unit recording (Tuleau-Malot et al., 2014; Reynaud-Bouret et al., 2014; Albert et al., 2015; 2016). These methods will require renewed efforts in order to be applied on massive recordings, as MEA and secondarily for single unit recording on freely moving animals. Other approaches will be pursued in order to develop new statistical methods for spike sorting and LFP/single unit recordings correlations. ComputaBrain will also collaborate with LJAD through the proposal “Factorisation de matrices pour le traitement de données biomédicales” submitted to the IDEX AAP Data Sciences, for spike sorting and neuronal graph reconstruction. At INRIA, deterministic models are developed from synapses to learning of sequences of items in memory, and multilevel statistical analysis is well established. Network dynamics is also an area of expertise, in close collaboration with LJAD researchers (equivariant and heteroclinic dynamics (Chossat et al., 2016), slow-fast systems (Desroches et al., 2016)). At BCL, human behavior and cognitive processes are investigated experimentally and modeled by biologically inspired networks of the cerebral cortex (Brunel & Lavigne, 2009; Lavigne et al., 2011). However, cognitive modeling still lacks statistical models of the data and also biological data on synaptic learning. The collaboration between partners aims at filling all these gaps. External collaborations with UCC and UofA will provide additional knowledge for the mathematical analysis of learning and processing of concepts. 

Impact

Based on a local integration between biological/statistical/computer science/cognition laboratories, scientific impacts consist of:
1. Provide a multilevel bio-analysis of the specific networks implicated in cognitive networks (extending current competences to network analysis),
2. Optimize the statistical analysis of biological data to enhance pertinence,
3. Development of biologically-realistic in silico models of synapse and neuron function,
4. Mathematize brain network-based learning.


These goals allow reaching the perspectives described by the work packages, opening new local opportunities to develop new machine learning algorithms, allow further electronic implementations, study brain diseases (as Alzheimer and epilepsy) and determine corresponding therapeutics, as well as providing new educational research-based advantages to UCA Jedi .

Members

Members1 members2

The main purpose of ComputaBrain is to bring together the local technical and human competences on learning & cognition and promote interdisciplinary work.

35 collaborators of 6 local institutes (IPMC, INRIA Sophia, I3S, IBV and BCL). 

Principal investigators: Alexandre Muzy (I3S) and Hélène Marie (IPMC).

Name

Structure

Role in ComputaBrain

Alexandre Muzy (AM)

I3S

Mathematical analysis of sequential machines and activity-based learning/simulation

Gilles Bernot (GB)

I3S

Validation of the results in formal methods (logics)

Frédéric Precioso (FP)

I3S

Validation of the results in deep learning

Goeffrey Portelli (GP)

I3S

Validation of the results in deep learning

Ophélie

Guinaudeau (OG)

I3S

Study of dendrite impact in activity measures

Cyrille Marscat (CM)

I3S

Neuron synchronization modeling

Hélène Marie (HM)

IPMC

Biological analysis of synaptic and neuronal plasticity

Paula Pousinha (PP)

IPMC

Biological analysis of synaptic and neuronal plasticity

Ingrid Bethus (IB)

IPMC

Biological learning and in vivo single unit activity

Massimo Mantegazza (MM)

IPMC

Biological analysis of neuronal excitability

Fabrice Duprat (FD)

IPMC

Biological analysis of neuronal excitability

Isabelle Lena (IL)

IPMC

Biological analysis of neuronal excitability and behavior

Michele Studer (MS)

IBV

Biological analysis of neuronal assemblies in ex vivo experiments

Alessandro Simi (AS)

IBV

Biological analysis of neuronal assemblies in ex vivo experiments

Karim Lounici (KL)

LJAD

Mathematical modeling of sequential learning

Franck Grammont (FG)

LJAD

Mathematical modeling of neuronal assemblies in ex vivo experiments

Patricia Reynaud-Bouret (PR)

LJAD

Mathematical modeling of sequential learning

Pascal Chossat (PC)

LJAD

Sequential activation of concepts based on heteroclinic chains

Romain Veltz (RV)

INRIA

Synaptic plasticity in silico modeling

Maureen Clerc (MC)

INRIA

Coherency of multiscale data

Theo Papadopoulo (TP)

INRIA

Coherency of multiscale data

Mathieu Desroches (MD)

INRIA

Excitability in silico modeling

Elif Köksal Ersöz (EK)

INRIA

Modeling of associative memory

Frédéric Lavigne (FL)

BCL

Human cognitive modeling

Fanny Meunier (FM)

BCL

Human cognitive modeling

Bernard Zeigler (BZ)

UofA*

Abstraction of dynamic systems structure and dynamics by morphisms

Maciej (Martin) Krupa (MK)

UCC*

Sequential activation on concepts based on heteroclinic chains

References

Albert M., Bouret Y., Fromont M., Reynaud-Bouret P. (2016) Surrogate data methods based on a shuffling of the trials for synchrony detection: the centering issue. Neural Computation, 28(1):2352–2392, 2016. doi: 10.1162/NECO a 00839. URL https://hal.archives-ouvertes.fr/hal-01154918.

Albert M., Bouret Y., Fromont M., Reynaud-Bouret P. (2015) Bootstrap and permutation tests of independence for point processes. The Annals of Statistics, 43(6):2537–2564. doi: 10.1214/15-AOS1351. URL https://hal.archives-ouvertes.fr/hal-01001984.

Alfano C., Magrinelli E., Harb K., Hevner R. F. and Studer M.  (2014) Postmitotic control of sensory area specification during neocortical development. Nature Communications, 5:5632.

Bethus I, Tse D, Morris RGM.  (2010) Dopamine and memory: modulation of the persistence of memory for novel hippocampal NMDA receptor-dependent paired-associates. J Neurosci., 30(5):1610-8.

Bianchi D, Marasco A, Limongiello A, Marchetti C, Marie H, Tirozzi B, Migliore M (2012) On the mechanisms underlying the depolarization block in the spiking dynamics of CA1 pyramidal neurons. Journal of Computational Neuroscience 33(2):207-225.  

Bianchi D, Michele PD, Marchetti C, Tirozzi B, Cuomo S, Marie H, Migliore M. (2014)  Effects of increasing CREB-dependent transcription on the storage and recall processes in a hippocampal CA1 microcircuit.  Hippocampus. 24(2):165-77.  

Brunel N. (2016) Is cortical connectivity optimized for storing information? Nat Neurosci., 19(5): 749-55.

Brunel, N. & Lavigne, F. (2009). Semantic Priming in a Cortical Network Model. The Journal of Cognitive Neuroscience, 21:12, 2300-2319.

Cestèle S., Schiavon E., Rusconi R., Franceschetti S. and Mantegazza M. (2013) A non-functional NaV1.1 FHM mutant transformed into gain of function by partial rescue of folding defects. PNAS USA. 110(43):17546-51.

Chin J., & Scharfman H.E (2013) Shared cognitive and behavioral impairments in epilepsy and Alzheimer’s disease and potential underlying mechanisms Epilepsy Behav. 26(3): 343–351

Chossat P, Krupa M, Lavigne F. (2016) Latching dynamics in neural networks with synaptic depression. arXiv:1611.03645v2.

Desroches M, Krupa M, Rodrigues S. (2016) Spike-adding in parabolic bursters: the role of folded-saddle canards. Phys. D 331(1): 58-70.

Devaine M., Gaillard P., Goude Y., Stoltz G. (2013) Forecasting electricity consumption by aggregating specialized experts. Machine Learning, Springer Verlag, 2013, 90 (2), pp.231-260.

Guerrini R., Marini C. and Mantegazza M. (2014) Genetic epilepsy syndromes without structural brain abnormalities: clinical features and experimental models. Neurotherapeutics 11(2):269-85.

Harb K., Magrinelli E., Nicolas C.S, Lukianets N., Frangeul L., Pietri M., Sun T., Sandoz G., Grammont F., Jabaudon D., Studer M.* and Alfano C.* (2016) Area-specific development of distinct neocortical neuron subclasses is regulated by postnatal epigenetic modifications. eLife, Jan 27;5.

Hedrich U.B.S, Liautard C., Kirschenbaum D., Pofahl M., Lavigne J., Liu Y, Theiss S., Slotta J.,  Escayg A., Dihné M., Beck H., Mantegazza M., Lerche H. (2014) Impaired action potential initiation in GABAergic interneurons causes hyperexcitable networks in an epileptic mouse model carrying a human NaV1.1 mutation. Journal of Neuroscience 35(45):14874-14889.

Herrmann M., Ruppin E., Usher M. (1993) A neural model of the dynamic activation of memory. Biological Cybernetics. 68: 455–463.

Kropff, E., Treves, A. (2007) The complexity of latching transitions in large scale cortical networks. Natural Computing. 6:169–185.


Lanté F, Chafai M, Raymond EF, Salgueiro Pereira AR, Mouska X, Kootar S,  Barik J, Bethus I,  Marie H. (2015) Subchronic glucocorticoid receptor inhibition rescues early episodic memory and synaptic plasticity deficits in a mouse model of Alzheimer’s disease. Neuropsychopharmacology. 40:1772-81.

Lavigne F, Avnaïm F and Dumercy L (2014) Inter-synaptic learning of combination rules in a cortical network model. Front. Psychol. 5:842. doi: 10.3389/fpsyg.2014.00842.

Lavigne, F., Dumercy, L. & Darmon, N. (2011) Determinants of Multiple Semantic Priming: A Meta-Analysis and Spike Frequency Adaptive Model of a Cortical Network. The Journal of Cognitive Neuroscience, Vol. 23, No. 6: 1447–1474.

Lerner, I., Bentin, S., Shriki, O. (2012a) Spreading activation in an attractor network with latching dynamics: automatic semantic priming revisited. Cogn. Sci., 36, 1339–1382.

Muzy, A., Zeigler, B.P. (2014a) Activity-based credit assignment heuristic for simulation-based stochastic search in a hierarchical model base of systems, IEEE Systems Journal, vol.PP, no.99, 1-12.

Muzy, A., Zeigler, B.P. (2014b) A conjecture from learning simulations of series and parallel connections of components, the 26th European Modeling and Simulation Symposium, Sept. 2014, Bordeaux, France, 550-557.

Muzy, A., Lerasle, M., Grammont, F., Dao, V.T., Hill, D. (2016) Parallel and pseudorandom discrete event system specification vs. networks of spiking neurons: Formalization and preliminary implementation results
The 2016 International Conference on High Performance Computing & Simulation (HPCS 2016), International Workshop on Parallel Computations for Neural Networks (PCNN 2016), Jul 2016, Innsbruck, Austria.

Muzy, A., Zeigler, B.P. (2016) Iterative specification of input-output dynamic systems and implications for spiky neuronal networks, submitted to IEEE Systems journal.

Liautard C., Scalmani P., Carriero G., de Curtis M., Franceschetti S. and Mantegazza M. (2013) Hippocampal hyperexcitability and specific epileptiform activity in a mouse model of Dravet syndrome. Epilepsia. 54(7):1251-1261

Marchetti C, Marie H. (2011) Hippocampal synaptic plasticity in Alzheimer's disease: what have we learned so far from transgenic models? Rev Neurosci. 22:373-402.

Marie H, Morishita W, Calakos N & Malenka RC (2005) Generation of Silent synapses by acute in vivo expression of active CaMKIV or CREB. Neuron 45:741-52.

Middei S, Ammassari-Teule M, Marie H. (2014) Synaptic plasticity under learning challenge. Neurobiology of Learning and Memory 115:108-15. 

Palop JJ, Mucke L. (2016)  Network abnormalities and interneuron dysfunction in Alzheimer disease. Nat Rev Neurosci. (12):777-792

Romani A, Marchetti C, Bianchi D, Leinekugel X, Poirazi P, Migliore M, Marie H (2013) Computational modeling of the effects of amyloid-beta on release probability at hippocampal synapses.  Front.  Comp. Neurosci. 7:1.

Tse D, Langston RF, Kakeyama M, Bethus I, Spooner PA, Wood ER, Witter MP, Morris RGM. (2007) Schemas and memory consolidation. Science,  316: 76-82.

Tuleau-Malot, C., Rouis, A., Grammont, F., Reynaud-Bouret, P. (2014) Multiple Tests based on a Gaussian Approximation of the Unitary Events method with delayed coincidence count. Neural Computation 1408–1454, July 2014. doi: 10.1162/NECO a 00604. URL https: //hal.archives-ouvertes.fr/hal-00757323.

Reynaud-Bouret, P., Rivoirard, V., Grammont, F., and Tuleau-Malot, C. (2014) Goodness-of-fit tests and nonparametric adaptive estimation for spike train analysis. Journal of Mathematical Neuroscience, page 4:3, April 2014. doi: 10.1186/2190-8567-4-3. URL https: //hal.archives-ouvertes.fr/hal-00789127.

Willem M, Tahirovic S, Busche MA, Ovsepian SV, Chafai M, Kootar S, Hornburg D, Evans LDB, Moore S, Daria A, Hampel H, Müller V, Giudici C, Nuscher B, Wenninger-Weinzierl A, Kremmer E, Heneka MT, Thal DR, Giedraitis V, Lannfelt L, Müller U, Livesey FJ, Meissner F, Herms J, Konnerth A, Marie H, Haass C. (2015) Secretase processing of APP inhibits neuronal activity in the hippocampus. Nature, 526(7573):443-7.

Zeigler, B.P., Muzy, A.M. (2016) Temporal Modeling of Neural Net Input/Output Behaviors: The Case of XOR, Second Generation General System Theory: Perspectives in Philosophy and Approaches in Complex Systems, MDPI, accepted for publication.

Zeigler, B.P., Kim, T., Praehofer, H. (2000) Theory of Modeling and Simulation, 2nd Edition, Elsevier.

style