Neural networks with transient state dynamics - Condensed Matter > Disordered Systems and Neural NetworksReport as inadecuate




Neural networks with transient state dynamics - Condensed Matter > Disordered Systems and Neural Networks - Download this document for free, or read online. Document in PDF available to download.

Abstract: We investigate dynamical systems characterized by a time series of distinctsemi-stable activity patterns, as they are observed in cortical neural activitypatterns. We propose and discuss a general mechanism allowing for an adiabaticcontinuation between attractor networks and a specific adjoined transient-statenetwork, which is strictly dissipative. Dynamical systems with transient statesretain functionality when their working point is autoregulated; avoidingprolonged periods of stasis or drifting into a regime of rapid fluctuations. Weshow, within a continuous-time neural network model, that a single localupdating rule for online learning allows simultaneously (i) for informationstorage via unsupervised Hebbian-type learning, (ii) for adaptive regulation ofthe working point and (iii) for the suppression of runaway synaptic growth.Simulation results are presented; the spontaneous breaking of time-reversalsymmetry and link symmetry are discussed.



Author: Claudius Gros

Source: https://arxiv.org/







Related documents