From spikes to rates, from manifolds to factors: the importance of low-dimensional dynamics in the brain.

Prof. Dr. Wulfram Gerstner, EPFL

TIME: Wednesday, Nov 5th, 2025 at 12:15 PM
WHERE: IMBIT NEXUS Lab, Georges-Köhler-Allee 201, 79110 Freiburg

ABSTRACT: There are a few basic well-known facts in neuroscience: The brain consists of MANY neurons; neurons communicate by spikes; spike trains look different between one trial and the next, suggesting a noisy process;  learned behavior such as arm movements is reliable brain dynamics despite the noise; neurons have receptive fields; populations of neurons are highly heterogeneous.   Morever,  more recently simultaneous recordings of hundreds of neurons from many labs have shown that the dynamics of brain activity can be visualized as dynamics moving in low-dimensional manifolds (e.g., a plane). And modeling work in many labs have shown that rate models work well to explain these dynamic phenomena.

In my talk I will try to connect these facts together using the concept of low-rank connectivity (Mastrogiuseppe and Ostojic) as the core of the theory. I will argue that the organization of receptive fields in visual cortex, the retrieval of memories, and the preparation of motor movements can all be described with the same class of models.   Surprisingly, this approach also provides an unexpected connection between spiking models and rate models.

Literature:

Mastrogiuseppe and Ostojic, NEURON, 2018 https://www.cell.com/neuron/fulltext/S0896-6273(18)30543-9;
Pezon et al.  bioRxiv (2024), https://www.biorxiv.org/content/10.1101/2024.02.28.582565v3.abstract

Biography: Wulfram Gerstner is Director of the Laboratory of Computational Neuroscience LCN at the EPFL. He studied physics at the universities of Tubingen and Munich and received a PhD from the Technical University of Munich. His research in computational neuroscience concentrates on models of spiking neurons, the dynamics of spiking neural networks and spike-timing dependent plasticity. More recently, he got interested in generalizations of Hebbian learning in the form of multi-factor learning rules and in the role of surprise for learning. He currently has a joint appointment at the School of Life Sciences and the School of Computer and Communications Sciences at the EPFL. He teaches courses for Physicists, Computer Scientists, Mathematicians, and Life Scientists. He is the recipient of the Valentino Braitenberg Award for Computational Neuroscience 2018 and a member of the Academy of Sciences and Literature Mainz (Germany).

The lecture will be followed by informal discussions with light snacks and drinks.

HOST: Christian Leibold