Connectionists: Western-Fields Seminar Series | Arvind Kumar

[Lyle Muller]

The seventh talk in the 2021 Western-Fields Seminar Series in Networks, Random Graphs, and Neuroscience<http://www.fields.utoronto.ca/activities/21-22/western-fields> is Thursday (30 September) at noon EDT.

Arvind Kumar<https://www.kth.se/profile/arvindku> (KTH Stockholm) will give a talk titled “Network structure and network activity dynamics: What really matters?” (abstract below). Dr. Kumar is a leader in computational neuroscience and the dynamics of spiking neural networks.

This seminar series features monthly virtual talks from a diverse group of researchers across machine learning, physics, and graph theory. Upcoming speakers include Alexander Lubotzky (Hebrew University of Jerusalem), Frances Skinner (Krembil Institute), and more.

Registration link: https://zoom.us/meeting/register/tJYuf-GppzkjHt0W5HMDpME2UpUiE7ntO5JS

—

Abstract: The wiring of neurons in the brain essentially determines the brain function. While the brain's wiring is not fixed, the overall structure of the connectivity is largely maintained over a long time. Therefore, there is a big interest in mapping the 'connectome' of the brains. But the key questions still remain unclear : Which properties of the network structure are crucial for network dynamics and/or function? How important is the network topology (small-world, scale-free etc.) and graph theoretic descriptors in shaping the network activity dynamics? To address these questions, we have compared the dynamics of different types of random, spatially connected and heterogeneous neuronal networks. We found that network topology by itself is a rather fragile concept and does not affect the network dynamics in a qualitative manner. By contrast, simple features of the network structure such as effective synaptic connectivity or excitation-inhibition balance and the shape of the connectivity kernel are more powerful in bringing about qualitative changes in the network activity. We also found that in spatial networks connection heterogeneities are only effective when they can create local imbalance of excitation and inhibition. This can be achieved only when connectivity inhomogeneities are correlated in physical (or connectivity) space. These results suggest mechanisms by which spatial distribution of heterogeneities can give rise to temporal sequences.  Finally, I will discuss to what extent we can see the signatures of neuron or synapse degeneration in the activity of a biological neural network.

--
Lyle Muller
http://mullerlab.ca

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20210927/3a8c2f94/attachment.html>