A Globally Asymptotically Stable Plasticity Rule for Firing Rate Homeostasis

P. Joshi and J. Triesch

Abstract:

How can neural circuits maintain stable activity states when they are constantly being modified by Hebbian processes that are notori- ous for being unstable? A new synaptic plasticity mechanism is presented here that enables a neuron to obtain homeostasis of its firing rate over longer timescales while leaving the neuron free to exhibit fluctuating dynamics in response to external inputs. Mathematical results demon- strate that this rule is globally asymptotically stable. Performance of the rule is benchmarked through simulations from single neuron to network level, using sigmoidal neurons as well as spiking neurons with dynamic synapses.

Reference: P. Joshi and J. Triesch. A globally asymptotically stable plasticity rule for firing rate homeostasis. In Artificial Neural Networks - ICANN 2008, 2008.