Abstract
We examine the conditions under which spike-timing-dependent plasticity (STDP) normalizes post-synaptic firing rates. Our simulations show that the rate normalization property of STDP is fragile and small changes in the LTD/LTP ratio or pre-synaptic input rates can lead to high firing rates. We propose an adaptive scheme to dynamically control the LTP/LTD ratio. The biophysics of synapses lead us to suggest a control mechanism using action potential-induced calcium influx, a known mediator of synaptic plasticity. This adaptive STDP rule is shown to stabilize the post-synaptic firing rates under a variety of perturbations.
Original language | English |
---|---|
Pages (from-to) | 189-194 |
Number of pages | 6 |
Journal | Neurocomputing |
Volume | 44-46 |
DOIs | |
State | Published - 2002 |
Keywords
- Hebbian
- Homeostatic mechanisms
- Learning
- Spike correlations