`Chaotic relaxation' in concurrently asynchronous neurodynamics

摘要

A mathematical framework for reconditioning additive-type modelsis proposed, and a neuro-operator, based on the chaotic relaxationparadigm, whose resulting dynamics is neither concurrently synchronousnor sequentially asynchronous is derived. Necessary and sufficientconditions guaranteeing concurrent asynchronous convergence areestablished in terms of contracting operators. Lyapunov exponents arealso computed to characterize the network dynamics and to ensure thatthroughput-limiting chaotic behavior in models reconditioned withconcurrently asynchronous algorithms has been eliminated