Computational Neuroscience Group

Our lab uses mathematical models of synapses, neurons and networks to explain aspects of perception and behaviour. We consider models for the cortical pyramidal neurons and micro-circuitries which are being experimentally investigated in vivo and in vitro at our Department.1 . A further interest is the neuronal substrate of learning and memory. One question we are addressing is how state-action sequences can be learned from an ongoing stream of synaptic inputs and a single delayed feedback signal2 . We also develop models of sensory processing and its interaction with cortical top-down signals. These explain experimental recordings from the visual cortex obtained while solving perceptual or classification tasks3 , 4 . Our models try to highlight some key mechanisms by which the interaction of synapses and neurons enables our brains to deal with everyday tasks.


  • New article in Elife: how natural-gradient descent enable efficient synaptic plasticity!
  • Kreutzer2022
    "Natural-gradient learning for spiking neurons".
    See Publications

  • From Cortical Microcircuits to Consciousness Symposium - Apr 11-13 - Info
  • One Talk and two Poster presentations:
    - Walter Senn
    : who also was the Scientific Chair of the symposium presented "Creative sleep, adversarial dreams, and the need of awareness."
    - Nicolas Deperrois: "Learning cortical representations through perturbed and adversarial dreaming."
    - Arno Granier: "The geometry of precision in predictive coding."


  • New contribution to NeurIPS: Learning with slow neurons is not a problem anymore!
    This work was selected for an oral presentation at the 35th Conference on Neural Information Processing Systems (NeurIPS 2021)!
  • LEnews
    "Latent Equilibrium: A unified learning theory for arbitrarily fast computation with arbitrarily slow neurons".
    See Publications - Press Release