Laboratory of Computational Neuroscience LCN - Brain Mind - I&C - Life Sciences - EPFL

Wulfram Gerstner - Research

Current research projects in my group cover the following areas:

• Temporal Coding in Networks of Spiking Neurons . At the LCN, as in many other labs, we have routinely used integrate-and-fire neurons and variants thereof to study temporal coding properties of spiking neurons. We have developed the Spike Response Model that emphasizes refractory properties of the neuron, as well as the adaptive exponential integrate-and-fire model (AdEx). Recently, we explored whether these neuron models can be used to predict neuronal activity in slice experiments on a spike-by-spike basis under random-current or random-conductance injection. Research in this area is persued by my PhD students and PostDocs Richard Naud, Christian Pozzorini, Felipe Gerhard, Skander Mensi Friedemann Zenke and Tim Vogels

  See also our book on Spiking Neuron Models as well as the recent papers of Badel et al. (2008) and Jolivet et al. (2006) or the short review from 2009 in Science

• Spike-timing dependent plasticity (STDP) and learning. I have been interest in the precise temporal conditions for synaptic plasticity since 1993 when we published a first paper in Biological Cybernetics. An STDP function with two phases (LTP for pre- before post and LTD for inverse timing) has been studied for the first time in a model of the barn owl auditory system - published in a Nature paper in 1996, slightly before the first experimental results on STDP appeared. In view of the wealth of experimental results, we have developed modern version of STDP models that emphasize triplets of spikes and the relation to BCM learning rules, as well as the relation to reinforcement learning. Currently, my PhD students and PostDocs Lorric Ziegler, Guillaume Hennequin, Friedemann Zenke and Tim Vogels continue this modeling effort.

  See our early papers in Biological Cybernetics (1993) and Nature (1996) as well as the recent articles of Pfister et al. (2006) in Neural Computation and Toyoizumi et al. (2005) in Proc. Natl. Acad. Sciences and Clopath et al. (2010) in Nature Neuroscience . See also the short review of STDP in Scholapedia as well as the recent work on Reward-modulated STDP in J. Neuroscience (2010)

• Link to Behavior and Psychophysics. On a different line of research, I have been interested for many years in how spatial representation by place cells can be implememted in a biologically plausible model. Many models only look at the problem of navigation, given an existing representation by a set of hippocampal place cells. Together with my former PhD-student Angelo Arleo we have developed a method to automatically create place fields with unsupervised growing networks while a robot moving around in an environment. My current PhD students Ricardo Chavariaga, Tomas Stroesslin, and Denis Sheynikhovich have persued this research further. In particular we have focused on questions of how the geometrical and non-geometrical features of the environment influence different navigational strategies. The model is learning using combinations of unsupervised Hebbian learning for the place cells and reinforcement learning for the choice of actions. See the papers of Arleo and Gerstner 2000 and Sheynikhovich et al. 2005 as well as the recent article of Sheynikhovich et al. 2009 in Psychological Reviews . The project is finished.

  Together with my colleague Carmen Sandi at EPFL I have co-supervised Gedi Luksys as a PhD studetns so as to explain behavioral learning data in rodents under different stress conditions with a reinforcement learning model. See Luksys, Gerstner, and Sandi (2009) in
  Nature Neuroscience 12:1180-1186
  

  Together with my colleague Olaf Blanke I co-supervise Danilo Jimenez Rezende to explore models that would explain the bodily self.

  Together with my colleague Michael Herzog we work on psychophysics of decision making with ultra-short stimuli.

Back to home page of Wulfram Gerstner.
Go to
LCN - Brain-Mind Institute - Computer and Communication Sciences - Life Sciences - EPFL