Computation in Neural Circuits
About
The group of Abigail Morrison in the Laboratory for Computational and Systems Neuroscience studies models of cognitive functions with a particular focus on implicit learning processes. Starting from assumptions about the computational function of a neural circuit, the group applies a top-down approach to develop biologically constrained spiking neural network models, thereby elucidating requirements on the circuit structure, dynamics, neuronal morphology and synaptic plasticity to realize the function in question. The group further constrains models through bottom-up approaches to uncover the effect of biological network features on the computational properties under investigation. A natural extension of such investigations is the transfer of neural computing principles to machine learning and neuromorphic computing paradigms.
Simulation studies involving plasticity require complex synapse models and are computationally expensive. The group therefore develops simulation techniques for massively parallel computers that are efficient in terms of both computation and memory consumption, and code generation methods for rapid prototyping of neuron and synapse models. See NEST.
Research Foci

Copyrights: A spiking neural network solving the Mountain Car problem self-organises to represent the task-relevant portions of the 2D input space (Weidel et al., 2021)
