Walter Senn

The Senn lab offers a Postdoc or PhD position on a computational model of cortical self-attention (supported by EBRAINS 2.0). Based on our current work and a collaboration with The Virtual Brain (Marmaduke Woodman), we seek for an implementation of neuronal self-attention mechanisms in thalamo-cortical circuits. The model is inspired by transformer-type architectures, but is consistent with experimentally observed cortical connectivity patterns. It shall be trained on cognitive tasks while being constrained to human cortical recordings. The Senn lab also offers a Postdoc or PhD position on the Neuronal Least-Action principle and its extension to long-term temporal processing and spikes (supported by the SNSF). The framework offers a rigorous description of the neuronal dynamics in cortical networks together with gradient-based synaptic learning rules. It will be applied to integrate-and-fire neurons with multiple intrinsic time constants. The project links to implementations in spike-based neuromorphic hardware.