Júlia Gallinaro: Associative properties of a structural plasticity rule based on firing rate homeostasis

BCCN Freiburg

The interaction between Hebbian and homeostatic plasticity in neuronal networks has recently received a lot of attention. Hebbian synaptic plasticity like STDP, known for its associative properties, leads to instabilities in recurrent networks, and different homeostatic mechanisms have been proposed to stabilize the learning process. While slow homeostatic plasticity has been observed in experiments, a mechanism fast enough to compensate instabilities on short time scales remains to be found. The goal of this work is to contribute another aspect to the understanding of this interaction: could associative properties also emerge from a rule based on homeostatic principles? We consider the maturation of networks in the primary visual cortex (V1) of mice as an example. In contrast to the situation right after eye-opening, neurons in adult V1 are more likely to connect to other neurons that have similar preferred orientations (PO). We simulate this maturation process in a recurrent network of leaky integrate-and-fire neurons, in which excitatory to excitatory connections are subject to a structural plasticity rule based on the homeostasis of firing rates. We found that upon stimulation that emulates early visual experience, the connection probability is indeed modulated according to the PO of neurons. Moreover, we could show that this effect is long-lasting and the emerging structure decays only slowly when the specific external stimulation is turned off. Our results demonstrate very clearly that associative properties can also emerge from a plasticity rule that is only based on firing rate homeostasis in single neurons, and that is not explicitly dependent on correlations between the activity of neurons.

Organized by

Henning Sprekeler

Location

TU Berlin, Room 4.063, Marchstraße 23, 10587 Berlin