Spike-timing dependent plasticity (STDP)

In standard Hebbian learning, a synaptic weight is increased if presynaptic and postsynaptic neuron are `simultaneously' active. If neurons communicate by spikes, the concept of simultaneity implies the pre- and postsynaptic spikes occur within some time window. Theory predicts that these time windows could have two phases corresponding to an increase (potentitiation) or decrease (depresseion) of the synaptic weight depending on the relative timing of pre- and postsynaptic spike. Such asymmetric learning rules with two phases have been found in recent experiments. We try to model such learning rules.

In this project we study neural network models with spiking neurons. We address, among others, the following questions:

  • Is there a simple learning rule that explains experiments under a variety of paradigms?
  • What are the computational advantages of spike-time dependent learning rules compared to rate-base learning?
  • Can we formulate efficient learning rules for temporal processes?
  • What would be `optimal' learning rules?
  • How can we include the action of neuromodulators or global feedback in the theory?

Collaborators:

Code TagTic.py

Previous collaborators:

  • Claudia Clopath
  • Eleni Vasilaki
  • Jean-Pascal Pfister
  • Taro Toyoizumi, LCN and Univ. of Tokyo

Some publications:

Clopath, Claudia ; Büsing, Lars ; Vasilaki, Eleni ; Gerstner, Wulfram (2010)
Connectivity reflects coding: a model of voltage-based STDP with homeostasis
Nature Neuroscience, vol. 13, num. 3, 2010, p. 344-352

Vasilaki, Eleni ; Frémaux, Nicolas ; Urbanczik, Robert ; Senn, Walter (2009)
Spike-Based Reinforcement Learning in Continuous State and Action Space: When Policy Gradient Methods Fail
PLoS Computational Biology, vol. 5, num. 12, 2009, p. e1000586

Clopath, Claudia ; Ziegler, Lorric ; Vasilaki, Eleni ; Büsing, Lars ; Gerstner, Wulfram (2008)
Tag-Trigger-Consolidation: A Model of Early and Late ong-Term-Potentiation and Depression
PLoS Comput Biol, vol. 4, num. 12, 2008, p. e1000248

J. P., Pfister, T. Toyoizumi, D. Barber, W. Gerstner (2006)
Optimal Spike-Timing Dependent Plasticity for Precise Action Potential Firing in Supervised Learning
Neural Computation, vol. 18, num. 6, 2006, p. 1309-1339

T. Toyoizumi, J.-P. Pfister, K. Aihara, and W. Gerstner (2005)
Generalized Bienenstock-Cooper-Munro rule for spiking neurons that maximizes information transmission
Proc. Natl. Acad. Sci. (USA), 102:5239-5244 [ pdf-file]

W. Gerstner and W. Kistler (2002).
Mathematical formulations of Hebbian Learning.
Biological Cybernetics, vol. 87, num. 5-6, 2002, p. 404-415

R. Kempter, W. Gerstner, and J. L. van Hemmen (1999)
Hebbian Learning and Spiking Neurons
Physical Review E, 59:4498-4514

Gerstner W, Kempter R, van Hemmen JL, and Wagner H (1996)
A neuronal learning rule for sub-millisecond temporal coding.
Nature, 383 :76-78

For additional references, consult the list of Publications of W. Gerstner.


Please send comments on this page to: wulfram.gerstner@epfl.ch