MC1 – Neurons and the Dynamics of Cognition: How Neurons Compute

Lecturer: Andreas Stöckel
Fields: Computational Neuroscience / Neuromorphic Computing

Content

While the brain does perform some sort of computation to produce cognition, it is clear that this sort of computation is wildly different from traditional computers, and indeed also wildly different from traditional machine learning neural networks. In this course, we identify the type of computation that biological neurons are good at (in particular, dynamical systems), and show how to build large-scale neural models that realize basic aspects of cognition (sensorimotor, memory, symbolic reasoning, action selection, learning, etc.). These models can either be made to be biologically realistic (to varying levels of detail) or mapped onto energy-efficient neuromorphic hardware.

Literature

  • Eliasmith, C. and Anderson, C. (2003). Neural engineering: Computation, representation, and dynamics in neurobiological systems. MIT Press, Cambridge, MA.
  • Eliasmith, C. et al., (2012). A large-scale model of the functioning brain. Science, 338:1202-1205.
  • Stöckel, A. et al., (2021). Connecting biological detail with neural computation: application to the cerebellar granule-golgi microcircuit. Topics in Cognitive Science, 13(3):515-533.
  • Dumont, N. S.-Y. et al., (2023) Biologically-based computation: How neural details and dynamics are suited for implementing a variety of algorithms. Brain Sciences, 13(2):245, Jan 2023.

Lecturer

Andreas Stöckel received his PhD in computer science at the University of Waterloo, Canada, in 2021. During his PhD, his research focused on integrating biological detail into the Neural Engineering Framework, a method for constructing large-scale models of neurobiological systems. His work specifically focused on harnessing nonlinear synaptic interactions and temporal tuning as computational resources. Today, he is a senior research scientist at Applied Brain Research Inc., where he co-designed the TSP1 time-series processor, a low-power neural-network accelerator chip that utilizes some of the techniques that he investigated during his PhD.

Affiliation: Applied Brain Research
Homepage: https://compneuro.uwaterloo.ca/people/andreas-stoeckel.html