Exploration of the dynamics of spiking neural networks / F. Rhéaume, DRDC Valcartier ; D. Grenier, Université Laval. : D68-6/005-2013-1E-PDF

"Liquid state machine is a technique that is well-suited for many spatiotemporal pattern recognition tasks found in defence applications. The technique exploits the complex dynamics of spiking neural networks. In general, Liquid state machines use biologically inspired neuron models such as the leaky-integrate-and-fire model. When structured into a network, leaky-integrate-and-fire neurons interact in a complex and a non-linear manner, making it difficult to understand their behavior in response to an input stimulus. In this memorandum, the response of simple neural systems to input stimulus is studied. The goal is to provide a first exploration of the dynamical state of spiking neurons in reaction to their stimulation by continuous inputs. For this purpose, small neural systems made of, respectively, one single neuron, three serial neurons, and a three-by-three network are analyzed"--Abstract, page i.

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.882747&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs