## NEURONAL DYNAMICS ACROSS MACROSCOPIC TIMESCALES

**Speaker:**Yue Liu, Boston University, Physics Department

**When:**September 14, 2020 (Mon), 02:00PM to 03:00PM (add to my calendar)

*This event is part of the PhD Final Oral Exams. *

The brain operates in a world with rich dynamics across a wide range of timescales, including those on the order of seconds and above. Behavioral experiments on memory and timing reveal striking similarities in the behavioral patterns across a vast range of timescales from seconds to minutes. To subserve these behavioral patterns and adapt to natural statistics, the collective activity of the large of number of neurons in the brain should exhibit similar dynamics over a spectrum of macroscopic timescales, an idea we refer to as scale-invariance. Recently, new techniques for large-scale and chronic measurements of neural activity open up the opportunity to investigate neuronal dynamics across different macroscopic timescales. This talk presents modeling, theoretical and empirical work that center around the idea of scale-invariance. After a brief introduction on the empirical evidence for scale-invariance, I will start with a biophysically-realistic neural circuit model that combines a detailed simulation of a calcium-activated membrane current with the mathematical formalism of the inverse Laplace transform to produce sequential neural activity with the scale-invariant property. Next, I will present a theoretical analysis on the ability of linear recurrent neural networks to generate scale-invariant neural activity. It is shown that the network connectivity matrix should have a geometric series of eigenvalues and translated eigenvectors if the eigenvalues are real and distinct. Lastly, I will show an empirical analysis of neural data motivated by scale-invariance. The analysis reveals the existence of repeatable neuronal dynamics on the timescale of both seconds and minutes in multiple neural recordings of rodents performing cognitive tasks.

Zoom information https://bostonu.zoom.us/j/93271109794?pwd=MXIrNVpSbVlGai9WQkpUbm0yK1grZz09Meeting ID: 932 7110 9794 Passcode: 550550