Postdoc and PhD position on Biological Deep Learning

Job description: A postdoc and PhD position devoted to the implementation of error-backpropagation-like algorithms by cortical microcircuits is available in the Senn lab (University of Bern, Switzerland).

The project theoretically explores how cortical pyramidal neurons with their dendrites and surrounding interneurons can be embedded into a local microcircuit to support error-correction learning. The cortical microcircuits are stacked to multi-layer recurrent networks to be used for deep learning. The goal is to find network architectures that can process ongoing streams of sensory inputs while feedback signaling and synaptic plasticity are continuously operating.

The project is part of a collaboration with Yoshua Bengio (University of Montreal) on deep learning and with Matthew Larkum (Humboldt University Berlin) on dendritic processing by cortical pyramidal neurons.

The research in the Senn lab is devoted to models of synaptic plasticity and dendritic integration in the context of learning and behavior. We have recently suggested to consider “Learning by the dendritic prediction of somatic spiking” (Neuron 2014) that assigns an intrinsic computational task to neurons and dendrites. This concept was extended to include dendritic nonlinearities that gives the dendritic tree the functionality of an error-backpropagating network (PLoS Comp Biol 2016, see also Computational Neuroscience Group ). It will now be further extended to cortical microcircuits and deep learning.

deal candidates have a strong background in machine learning, computational neuroscience and applied mathematics. Please send CV, publication list, letter of motivation and addresses for 3 reference letters to Prof. Dr.phil.nat. Walter Senn ( ) and Sabine Herzog Bochud ( ).