Winnerless Competition and the olfactory system of insects
As everybody agrees time is playing a key role in neural information processing. Even if some neural system is able to process the information "algebraically", without time, by "identity" or "spatial" coding, real neural systems include time in the encoding space. Spatio-temporal encoding of neural information has many advantages for higher level processing (recognition, binding, and decision making for the behavior). In particular, including time in the information processing helps to solve the fundamental contradiction between robustness and sensitivity (stability-sensitivity dilemma).
We are working in a new direction in spatio-temporal information processing based on experiments performed in the locust olfactory system (collaboration with Gilles Laurent lab,Caltech.), the orientation sensory system of the marine mollusk Clione and hippocampal place cell networks. Our new concept uses the transformation of incoming spatial or identity information into spatio-temporal output based on the intrinsic switching dynamics of an inhibitory neural network with non-symmetric synaptic connections. This is called the Winner-Less Competition Principle (WLC). A network based on the WLC principle does not process with attractors but with separatrices, which are the mathematical basis for the observed switching of active neuron groups.
We obtained rigorous results about the stability of sequential switching in the framework of a sufficiently general model. This type of networks may solve many different functional tasks. Computation with separatrices can be an optimal principle for the design of new paradigms of artificial neural networks.
PeopleM. I. Rabinovich,
H. D. I. Abarbanel
M. I. Rabinovich Spatio-Temporal Neural Coding and WLC Principle: From Experiments to Understanding (powerpoint presentation)
P. Varona, M. I. Rabinovich, A. I. Selverston and Y. I. Arshavsky
Competition between sensory neurons generates chaos: A possible mechanism for
molluscan hunting behavior
M. Rabinovich, A. Volkovskii, P. Lecanda, R. Huerta,
H. D. I. Abarbanel, and G. Laurent Dynamical Encoding by Networks of Competing Neuron Groups: Winnerless
M. Bazhenov, M. Stopfer, M. I. Rabinovich, R. Huerta, H. D. I. Abarbanel,
T. J. Sejnowski, G. Laurent Model of transient oscillatory
synchronization in the locust antennal lobe.
M. Bazhenov, M. Stopfer, M. I. Rabinovich, H. D. I. Abarbanel,
T. J. Sejnowski, G. Laurent, Model of cellular and network
mechanisms for odor-evoked temporal patterning in the locust antennal
Seliger, Lev S. Tsimring, and Mikhail I. Rabinovich Dynamics-based
sequential memory: Winnerless competition of patterns
Gilles Laurent, Mark Stopfer, Rainer W Friedrich, Misha I Rabinovich,
Alexander Volkovskii, and Henry DI Abarbanel Odor Encoding as an Active,
Dynamical Process: Experiments, Computation, and Theory
M. I. Rabinovich, R. Huerta, A. Volkovskii, H. D. I. Abarbanel, M. Stopfer, G. Laurent Dynamical coding of sensory information with competitive networks JOURNAL OF PHYSIOLOGY-PARIS, SEP-DEC, 2000, V94(N5-6):465-471
|Comments? Contact||Terry Peters, Phone +1-858-534-7753, tpeters (at) ucsd.edu|