Zhigulin, Valentin P. (2004) Multiple-scale dynamics in neural systems: learning, synchronization and network oscillations. Dissertation (Ph.D.), California Institute of Technology. http://resolver.caltech.edu/CaltechETD:etd-05132004-175650
Many dynamical processes that take place in neural systems involve interactions between multiple temporal and/or spatial scales which lead to the emergence of new dynamical phenomena. Two of them are studied in this thesis: learning-induced robustness and enhancement of synchronization in small neural circuits; and emergence of global spatio-temporal dynamics from local interactions in neural networks.
Chapter 2 presents the study of synchronization of two model neurons coupled through a synapse with spike-timing dependent plasticity (STDP). It shows that this form of learning leads to the enlargement of frequency locking zones and makes synchronization much more robust to noise than classical synchronization mediated by non-plastic synapses. A simple discrete-time map model is presented that enables deep understanding of this phenomenon and demonstrates its generality. Chapter 3 extends these results by demonstrating enhancement of synchronization in a hybrid circuit with living postsynaptic neuron. The robustness of STDP-mediated synchronization is further confirmed with simulations of stochastic plasticity.
Chapter 4 studies the entrainment of a heterogeneous network of electrically coupled neurons by periodic stimulation. It demonstrates that, when compared to the case of non-plastic input synapses, inputs with STDP enhance coherence of network oscillations and improve robustness of synchronization to the variability of network properties. The observed mechanism may play a role in synchronization of hippocampal neural ensembles.
Chapter 5 proposes a new type of artificial synaptic connection that combines fast reaction of an electrical synapse with plasticity of a chemical synapse. It shows that such synapse mediates regularization of chaos in a circuit of two chaotic bursting neurons and leads to structural stability of the regularized state. Such plastic electrical synapse may be used in the development of robust neural prosthetics.
Chapter 6 suggests a new approach to the study of spatio-temporal network dynamics. The approach is based on the analysis of dynamical motifs - small subnetworks with periodic and chaotic dynamics. It is used to explain the transition from quiescence to periodic and chaotic dynamics in simulations of randomly connected neural networks and the domination of periodic dynamics in simulations of spatially distributed networks with local connectivity.
|Item Type:||Thesis (Dissertation (Ph.D.))|
|Subject Keywords:||hebbian learning; nonlinear neurodynamics; synaptic plasticity|
|Degree Grantor:||California Institute of Technology|
|Division:||Physics, Mathematics and Astronomy|
|Thesis Availability:||Public (worldwide access)|
|Defense Date:||12 May 2004|
|Non-Caltech Author Email:||zhigulin (AT) alumni.caltech.edu|
|Default Usage Policy:||No commercial reproduction, distribution, display or performance rights in this work are provided.|
|Deposited By:||Imported from ETD-db|
|Deposited On:||28 May 2004|
|Last Modified:||26 Dec 2012 02:41|
- Final Version
See Usage Policy.
Repository Staff Only: item control page