You are here: Home Talks & Events Bernstein Seminar 2011 Matthieu Gilson (Lab for Neural Circuit Theory, RIKEN Brain Science Institute, Hirosawa, Wako-shi, Saitama, Japan)

Matthieu Gilson (Lab for Neural Circuit Theory, RIKEN Brain Science Institute, Hirosawa, Wako-shi, Saitama, Japan)

— filed under:

"Pattern spiking activity and STDP: learning and detection" / Monday, May 9, 2011, 17:15 h

  • Bernstein Seminar
When May 09, 2011
from 05:15 PM to 07:20 PM
Where Lecture Hall, Hansastr. 9a
Add event to calendar vCal









The Bernstein Center Freiburg

Bernstein Seminar
Matthieu Gilson
Lab for Neural Circuit Theory
RIKEN Brain Science Institute
Hirosawa, Wako-shi, Saitama, Japan

Pattern spiking activity and STDP:
learning and detection

Monday, May 9, 2011

17:15 h

Lecture Hall (ground floor)
Bernstein Center Freiburg
Hansastraße 9A
79104 Freiburg
Spike-timing-dependent plasticity (STDP) is hypothesised to tune neuronal circuitry by modifying the synaptic strengths depending on the precise timing of spiking activity. Many roles for STDP have been proposed in theoretical and experimental studies over the past decade, including detection of millisecond-precise spike sequences, independent component analysis, formation of cell assemblies and self-organization of ongoing activity. However, it is still unclear what type of spiking information can be captured by STDP, in particular for in-vivo activity. Here we examine the learning and subsequent detection of spiking pattern activity using a mathematical analysis. We focus on two models of pattern based on spike coordination (spike pattern) and fast modulations of instantaneous firing rates that mimic peristimulus time histograms (PSTH), respectively. This allows us to consider different types of variability in the input spike trains: spike count vs. temporal precision, and individual vs. collective properties. We show how, under general assumptions, pattern activity induces spike-time cross-correlations that dominate the learning process. This means that the collective temporal structure of the input spike trains matters, rather than the variability for individual spike trains (e.g., Fano factor). Altogether, a (noisy) Poisson neuron can be trained to robustly learn and detect both spike and PSTH-like patterns with temporal variability up to 20 ms, performing almost as well as a leaky integrate-and-fire neuron.
The talk is open to the public. Guests are cordially invited!
Personal tools