WINNERLESS COMPETITION PRINCIPLE IN NEUROSCIENCE Mikhail Rabinovich INLS University of California, San Diego ’

Slides:



Advertisements
Similar presentations
Introduction to Neural Networks
Advertisements

On Bubbles and Drifts: Continuous attractor networks in brain models
Introduction to Neural Networks Computing
DYNAMICS OF RANDOM BOOLEAN NETWORKS James F. Lynch Clarkson University.
Phase Bursting Rhythms in Inhibitory Rings Matthew Brooks, Robert Clewley, and Andrey Shilnikov Abstract Leech Heart Interneuron Model 3-Cell Inhibitory.
Neural Network of the Cerebellum: Temporal Discrimination and the Timing of Responses Michael D. Mauk Dean V. Buonomano.
Lecture 12: olfaction: the insect antennal lobe References: H C Mulvad, thesis ( Ch 2http://
Gain control in insect olfaction for efficient odor recognition Ramón Huerta Institute for Nonlinear Science UCSD.
A model for spatio-temporal odor representation in the locust antennal lobe Experimental results (in vivo recordings from locust) Model of the antennal.
5/16/2015Intelligent Systems and Soft Computing1 Introduction Introduction Hebbian learning Hebbian learning Generalised Hebbian learning algorithm Generalised.
Artificial neural networks:
Synchrony in Neural Systems: a very brief, biased, basic view Tim Lewis UC Davis NIMBIOS Workshop on Synchrony April 11, 2011.
Part II: Population Models BOOK: Spiking Neuron Models, W. Gerstner and W. Kistler Cambridge University Press, 2002 Chapters 6-9 Laboratory of Computational.
Introduction to the mathematical modeling of neuronal networks Amitabha Bose Jawaharlal Nehru University & New Jersey Institute of Technology IISER, Pune.
FAPs Bat avoidance is modulated by a constant stream of sensory input.
CH12: Neural Synchrony James Sulzer Background Stability and steady states of neural firing, phase plane analysis (CH6) Firing Dynamics (CH7)
[1].Edward G. Jones, Microcolumns in the Cerebral Cortex, Proc. of National Academy of Science of United States of America, vol. 97(10), 2000, pp
Michale Fee McGovern Institute for Brain Research Department of Brain and Cognitive Sciences MIT Jerusalem in Motion Workshop Jerusalem, Israel December.
Stable Propagation of Synchronous Spiking in Cortical Neural Networks Markus Diesmann, Marc-Oliver Gewaltig, Ad Aertsen Nature 402: Flavio Frohlich.
Dynamical Systems Analysis III: Phase Portraits By Peter Woolf University of Michigan Michigan Chemical Process Dynamics and Controls.
Levels in Computational Neuroscience Reasonably good understanding (for our purposes!) Poor understanding Poorer understanding Very poorer understanding.
How does the mind process all the information it receives?
A globally asymptotically stable plasticity rule for firing rate homeostasis Prashant Joshi & Jochen Triesch
How facilitation influences an attractor model of decision making Larissa Albantakis.
Laurent Itti: CS599 – Computational Architectures in Biological Vision, USC Lecture 7: Coding and Representation 1 Computational Architectures in.
Itti: CS564 - Brain Theory and Artificial Intelligence. Systems Concepts 1 CS564 - Brain Theory and Artificial Intelligence University of Southern California.
Connected Populations: oscillations, competition and spatial continuum (field equations) Lecture 12 Course: Neural Networks and Biological Modeling Wulfram.
Reflex Physiology. Reflex Arc The reflex arc governs the operation of reflexes. Nerve impulses follow nerve pathways as they travel through the nervous.
1 On Bubbles and Drifts: Continuous attractor networks and their relation to working memory, path integration, population decoding, attention, and motor.
THE ROLE OF NEURONS IN PERCEPTION Basic Question How can the messages sent by neurons represent objects in the environment?
Chaos in Neural Network Theme presentation Cui, Shuoyang 03/08/2005.
Dynamical Encoding by Networks of Competing Neuron Groups : Winnerless Competition M. Rabinovich 1, A. Volkovskii 1, P. Lecanda 2,3, R. Huerta 1,2, H.D.I.
Biological Modeling of Neural Networks: Week 14 – Dynamics and Plasticity Wulfram Gerstner EPFL, Lausanne, Switzerland 14.1 Reservoir computing - Complex.
Lecture 10: Mean Field theory with fluctuations and correlations Reference: A Lerchner et al, Response Variability in Balanced Cortical Networks, q-bio.NC/ ,
Cognition, Brain and Consciousness: An Introduction to Cognitive Neuroscience Edited by Bernard J. Baars and Nicole M. Gage 2007 Academic Press Chapter.
Dynamical network motifs: building blocks of complex dynamics in biological networks Valentin Zhigulin Department of Physics, Caltech, and Institute for.
Rhythmic Movements Questions: –How do they happen? –What do they mean? –Where do they come from? Reflex chain? Sequential pattern of activation? Reverberatory.
John Wordsworth, Peter Ashwin, Gabor Orosz, Stuart Townley Mathematics Research Institute University of Exeter.
Biomedical Sciences BI20B2 Sensory Systems Human Physiology - The basis of medicine Pocock & Richards,Chapter 8 Human Physiology - An integrated approach.
Review of the CPG How do we know the circuit pattern: Who connects to who? Excitatory vs inhibitory connections? How do we know a neuron is part of a CPG.
Chapter 7. Network models Firing rate model for neuron as a simplification for network analysis Neural coordinate transformation as an example of feed-forward.
Multiple attractors and transient synchrony in a model for an insect's antennal lobe Joint work with B. Smith, W. Just and S. Ahn.
The Function of Synchrony Marieke Rohde Reading Group DyStURB (Dynamical Structures to Understand Real Brains)
Modeling Neural Networks Christopher Krycho Advisor: Dr. Eric Abraham May 14, 2009.
Lecture 21 Neural Modeling II Martin Giese. Aim of this Class Account for experimentally observed effects in motion perception with the simple neuronal.
Synchronization in complex network topologies
Biological Neural Network & Nonlinear Dynamics Biological Neural Network Similar Neural Network to Real Neural Networks Membrane Potential Potential of.
Biological Modeling of Neural Networks Week 4 – Reducing detail - Adding detail Wulfram Gerstner EPFL, Lausanne, Switzerland 3.1 From Hodgkin-Huxley to.
Syntactic Language Processing through Hierarchical Heteroclinic Networks Workshop on Heteroclinic Dynamics in Neuroscience University of Nice December,
Neural Synchronization via Potassium Signaling. Neurons Interactions Neurons can communicate with each other via chemical or electrical synapses. Chemical.
Information encoding and processing via spatio-temporal spike patterns in cortical networks Misha Tsodyks, Dept of Neurobiology, Weizmann Institute, Rehovot,
Ch 9. Rhythms and Synchrony 9.7 Adaptive Cooperative Systems, Martin Beckerman, Summarized by M.-O. Heo Biointelligence Laboratory, Seoul National.
Biological Modeling of Neural Networks: Week 15 – Fast Transients and Rate models Wulfram Gerstner EPFL, Lausanne, Switzerland 15.1 Review Populations.
From LIF to HH Equivalent circuit for passive membrane The Hodgkin-Huxley model for active membrane Analysis of excitability and refractoriness using the.
Network Models (2) LECTURE 7. I.Introduction − Basic concepts of neural networks II.Realistic neural networks − Homogeneous excitatory and inhibitory.
Where are we? What’s left? HW 7 due on Wednesday Finish learning this week. Exam #4 next Monday Final Exam is a take-home handed out next Friday in class.
Biological Modeling of Neural Networks Week 4 Reducing detail: Analysis of 2D models Wulfram Gerstner EPFL, Lausanne, Switzerland 3.1 From Hodgkin-Huxley.
J. Kubalík, Gerstner Laboratory for Intelligent Decision Making and Control Artificial Neural Networks II - Outline Cascade Nets and Cascade-Correlation.
ZAS Tandem Workshop December, 11 – 13, 2010 Peter beim Graben
How Neurons Do Integrals
9. Continuous attractor and competitive networks
Collins Assisi, Mark Stopfer, Maxim Bazhenov  Neuron 
OCNC Statistical Approach to Neural Learning and Population Coding ---- Introduction to Mathematical.
Volume 30, Issue 2, Pages (May 2001)
Volume 36, Issue 5, Pages (December 2002)
Neuronal Decision-Making Circuits
Collins Assisi, Mark Stopfer, Maxim Bazhenov  Neuron 
Information Processing by Neuronal Populations Chapter 5 Measuring distributed properties of neural representations beyond the decoding of local variables:
Volume 30, Issue 2, Pages (May 2001)
Continuous attractor neural networks (CANNs)
Presentation transcript:

WINNERLESS COMPETITION PRINCIPLE IN NEUROSCIENCE Mikhail Rabinovich INLS University of California, San Diego ’

competition stimulus Winnerless without + dependent = Competition WINNER clique Principle

Hierarchy of the Models  Network with realistic H-H model neurons & random inhibitory & excitatory connections  Network with FitzHugh-Nagumo spiking neurons  Lotka-Volterra type model to describe the spiking rate of the Principal Neurons (PNs)

From standard rate equations to Lotka-Volterra type model

Stimulus dependent Rate Model is the strength of excitation in i by k is the excitation from the other neural ensembles is an external action is the strength of inhibition in i by j Is the firing rate of neuron i

Canonical L-V model (N>3) A heteroclinic sequence consists of finitely many saddle equilibria and finitely many separatrices connecting these equilibria. The heteroclinic sequence can serve as an attracting set if every saddle point has only one unstable direction. The condition for this is: Necessary condition for stability: i+1 i

Canonical Lotka-Volterra model Rigorous results (N=3) Then the heteroclinic contour is a global attractor if A noise transfer the heteroclinic contour to a stable limit cycle with the same order of a sequential switching Consider the matrix

WLC Principle & SHS (rate model) Geometrical image of the switching activity in the phase space is the orbit in the vicinity of the heteroclinic sequence Geometrical image of the switching activity in the phase space is the orbit in the vicinity of the heteroclinic sequence

WLC Principle & SHS (H-H neurons) Geometrical image of the switching activity in the phase space is the orbit in the vicinity of the heteroclinic contour Geometrical image of the switching activity in the phase space is the orbit in the vicinity of the heteroclinic contour

WLC in a network of three spiking-bursting neurons

The main questions: The main questions:  How does sensory information transform into behavior in a robust and reproducible way?  Do neural systems generate new information based on their sensory inputs?  Can transient dynamics be reproducible?

WLC dynamics of the piloric CPG: experiment & theory

Real time Clione’s hunting behavior

Clione’s hunting behavior

Clione’s neural circuit

WLC can generate an irregular but reproducible sequence   All connections are inhibitory   The SRCs are asymmetrically connected   There is 30% connectivity among the neurons   The hunting neuron excites allSCHs at variable strength Model assumptions

Projection of the strange attractor from the 6D phase space of the statocyst network

Weak reciprocal excitation stabilizes WLC dynamics: Birth of the stable limit cycle in the vicinity of the former heteroclinic sequence

Conductance-based model for “Winner take all” and “Winnerless” competition Winnerless Winner take all

Sequential dynamics of statocyst neurons

Motor output dynamics Firing rates of 4 different tail motorneurons at different burst episodes In spite of the irregularity the sequence is preserved

IMAGES OF THE DYNAMICAL SEQUENCES

Spatio-temporal coding in the Antennal Lobe of Locust (space = odor space) Spatio-temporal coding in the Antennal Lobe of Locust (space = odor space) Lessons from the experiments: The key role of the inhibition Nonsymmetric connections No direct connection between PNs

Time inputoutput Transformation of the identity input Into spatio-temporal input Into spatio-temporal output based on the intrinsic sequential dynamics of the neural ensemble Winnerless Competition Principle & New Dynamical Object: Stable Heteroclinic Sequence WLC & SHS

Transient dynamics of the bee antennal lobe activity during post-stimulus relaxation

Low dimensional projection of Trajectories Representing PN Population Response over Time

Stable Heteroclinic Sequence

Reproducible sequences in complex networks Inequalities for reproducibility:

Reproducibility of the heteroclinic sequence Neuron

Stable manifolds of the saddle points keep the divergent directions in check in the vicinity of a heteroclinic sequence

WLC in complex neural ensembles Complex network = many elements + + disordered connections + disordered connections Most important phenomena in complex systems on the edge of reproducibility are: (i) clustering, and (i) clustering, and (ii) competition (ii) competition

Rate model of the Random network   Is the step function

TWO REGIMES: A) B)

What controls the dynamics?

Phase portrait of the sequential activity

Chaos in random network

Reproducible transient sequence generated in random network

Reproducibility of the transient dynamics

Example of sequence

The network of songbird brain

HVC Songbird patterns HVC Songbird patterns

Self-organized WLC in a network with Hebbian learning

WLC in the network with local learning

WLC networks cooperation: * synchronization (i) electrical connections, (ii) synaptic connections; (iii) ultra-subharmonic synchronization ** competition

Synchronization of the CPGs of two different animals

Heteroclinic synchronization: Ultra-subharmonic locking

Heteroclinic Arnold tongues

Chaos between stairs of synchronizaton

Heteroclinic synchronization: Map’s description

Competition between learned sequences: on line decision making

The main messages:  The WLC principle & SHS do not depend on the level of the neuron & synapse description and can be realized by many different kinds of network architectures.  The WLC principle is able to solve a fundamental contradiction between robustness & sensitivity.  The transient sequence can be reproducible.  SHS can interact with each others: compete, synchronized & generate chaos. synchronized & generate chaos.

Thanks to the collaborators Thanks to the collaborators Valentin Afraimovich, Rafael Levi, Allan Selverston, Valentin Zhigulin, Henry Abarbanel, Yuri Arshavskii & Gilles Laurent

Spatio-temporal patterns in Clione’s nerves

WLC: Dynamics of the H-H network time (ms) Neuron

Reproducibility of the dynamics } – 10 trials time

Stimulation of statocyst nerve triggers a dynamical response in the motor neurons Motor output electro- physiological recording Motor output firing rates

Statocyst receptor activity during hunting episodes   The constant statocyst receptor activity turns into bursting in physostigmine   The activity is variable between episodes   A single receptor is active during different phases of the hunting episodes