[WTI-trainee] WTI Distinguished Speaker Series: Guillaume Lajoie

[Lafferty, John]

We’re pleased to announce that Professor Guillaume Lajoie will be speaking Thursday on “How end-to-end optimization and deep learning reveal computational roles of biological neurons’ diversity and adaptation’’ as part of the Distinguished Speaker Series of the Wu Tsai Institute at Yale University.

The talk will be held at 11:30am on Thursday, November 10 in William L. Harkness Hall, Room 117.  To arrange to meet with the speaker, please contact Jennifer Burt  (jennifer.burt at yale.edu)


[cid:91ad8a46-f809-4a88-a5f4-21ac8dfdedfc]

Abstract:

In this talk, I will present a novel method that leverages goal-driven optimization to reveal computational principles of biological neural dynamics, and how to exploit these principles to improve artificial systems. Specifically, I will focus on single-neuron dynamics which have rich and adaptive input-output properties. Features such as diverse f-I curves and spike frequency adaptation are known to place single neurons in optimal coding regimes when facing changing stimuli. Yet, it is still unclear how brain circuits exploit single neuron flexibility, and how network-level requirements may have shaped such cellular function. To answer this question, a multi-scaled approach is needed where the computations of single neurons and of neural circuits must be considered as a complete system. In this work, we use artificial neural networks to systematically investigate single neuron input-output adaptive mechanisms, optimized in an end-to-end fashion. Throughout the optimization process, each neuron has the liberty to modify its nonlinear activation function, parametrized to mimic f-I curves of biological neurons, and to learn adaptation strategies to modify activation functions in real time during a task. We find that such networks show much-improved robustness to noise and changes in input statistics. Importantly, we find that this procedure recovers precise coding strategies found in biological neurons, such as gain scaling and fractional order differentiation/integration, even though the optimization process could have converged to any other mechanism. Using tools from dynamical systems theory, we analyze the role of these emergent single-neuron properties and argue that neural diversity and adaptation play an active regularization role that enables neural circuits to optimally propagate information across time.

Bio:
Dr. Lajoie holds a PhD in applied mathematics from the University of Washington (UW), in Seattle WA, US. Subsequently, he held Bernstein Fellow position at the Max Planck Institue for Dynamics and Self-organization, in Goettingen, Germany, and a Washington Foundation Innovation Fellow position at the UW Center for Neuroengineering, Seattle, WA, US. He is an assistant professor at the mathematics department of the Université de Montréal since 2018, as well as a core academic member of Mila, the Quebec AI institute. Dr. Lajoie holds a Canada CIFAR AI research chair, as well as a Canada Research Chair in Neural Computation and Interfacing. His research group works at the intersection of AI and Neuroscience, or Neuro-AI, developing tools to better understand neural networks (biological or artificial) as well as algorithms for brain-machine interfaces for scientific and clinical use.

I am an applied mathematician interested in the interactions and commonalities of biological and artificial neural computations. My research group works at the intersection of AI and Neuroscience, or Neuro-AI, developing tools to better understand neural networks as well as algorithms for brain-machine interfaces for scientific and clinical use. My work is motivated by the remarkable ability of neural networks (biological and artificial) to learn and support complex, emergent computations. I use tools from dynamical systems, information theory, statistics and machine learning to address a range of problems, in collaboration with experimental neuroscientists and machine intelligence researchers.












-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.yale.edu/mailman/private/wti-trainee/attachments/20221107/b0497545/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: lajoie-flyer.png
Type: image/png
Size: 402884 bytes
Desc: lajoie-flyer.png
URL: <https://mailman.yale.edu/mailman/private/wti-trainee/attachments/20221107/b0497545/attachment-0001.png>