Research in the Computational Neuroscience Lab

The lab is broadly interested in the field of neuroscience, especially insofar as it addresses the questions of learning and memory. Learning is thought to change the connections between the neurons in the brain, a process called synaptic plasticity. Using mathematical and computational tools, we model synaptic plasticity across different time scales that reproduces experimental findings. We then study the role of synaptic plasticity, by constructing networks of artificial neurons with plastic synapses. We are working to tight collaboration with experimental laboratories, which measure connectivity changes and behavioral learning.

Below is a list of the research questions and projects of the lab:

  • We model long-term plasticity that reproduces slice experiments in a number of different systems (hippocampus, visual cortex, somatosensory cortex).

  • We model very long-term plasticity (memory over whole lifespans), which involves the mechanism of synapse consolidation.

  • We study the role of synaptic plasticity in receptive field development and in selectivity in the inputs. In particular, we study the mechanism of how learning can develop orientation selectivity in visual cortex (using dynamical systems techniques) and perform blind source separation (e.g. demixing sounds at a cocktail party).

  • We study recurrent artificial neural networks with plastic synapses. In particular, we study the close relation between the type of coding and the type of connectivity found in different brain areas. Among others, we could explain the recently observed differences between the connectivity in visual cortex and in barrel cortex. The theoretical work predicted that neurons with similar receptive fields have a high probability of being reciprocally connected, a finding that was later confirmed experimentally.

  • We study the functional implication of inhibitory plasticity, a regulatory mechanism to keep balance between inhibition and excitation.

  • Expanding to different brain areas, namely the cerebellum, we study how optimal memory capacity can predict synaptic weight distributions between Parallel Fibers and Purkinje Cells, using statistical physics methods.

  • We develop phenomenological models of cerebellar learning, reproducing learning behavior of wild-type and knock-out mice, as well as Purkinje cell electrophysiological measurements before and after learning.