Bulletin of the American Physical Society
Mid-Atlantic Section Meeting 2021
Volume 66, Number 18
Friday–Sunday, December 3–5, 2021; Rutgers University, New Brunswick, New Jersey
Session D04: Biophysics II |
Hide Abstracts |
Chair: Hanna Salman, University of Pittsburgh Room: 202B |
Saturday, December 4, 2021 11:15AM - 11:51AM |
D04.00001: Neural Optimal Feedback Control with Local Learning Rules Invited Speaker: Anirvan Sengupta A major problem in motor control is understanding how the brain plans and executes proper movements in the face of delayed and noisy stimuli. A prominent framework for addressing such control problems is Optimal Feedback Control (OFC). OFC generates control actions that optimize behaviorally relevant criteria by integrating noisy sensory stimuli and the predictions of an internal model using the Kalman filter or its extensions. However, a satisfactory neural model of Kalman filtering and control is lacking because existing proposals have the following limitations: not considering the delay of sensory feedback, training in alternating phases, and requiring knowledge of the noise covariance matrices, as well as that of systems dynamics. Moreover, many of these studies considered Kalman filtering in isolation, and not jointly with control. To address these shortcomings, we introduce a biologically plausible neural network, with local synaptic plasticity rules, which combines adaptive Kalman filtering with a model free control approach (i.e., policy gradient algorithm). This network performs system identification and Kalman filtering, without the need for multiple phases with distinct update rules or the knowledge of the noise covariances. It can perform state estimation with delayed sensory feedback, with the help of an internal model. It learns the control policy without requiring any knowledge of the dynamics, thus avoiding the need for weight transport. In this way, our implementation of OFC solves the credit assignment problem needed to produce the appropriate sensory-motor control in the presence of stimulus delay. [Preview Abstract] |
Saturday, December 4, 2021 11:51AM - 12:27PM |
D04.00002: Formation of neural connectivity: nature versus nurture Invited Speaker: Alexi Koulakov Neural development leads to the establishment of precise connectivity in the nervous system based on information contained in the genome. By contrasting the information capacities of cortical connectivity and the genome, I will argue that simplifying rules are necessary in order to create cortical connections from the limited set of instructions. Such rules contain compact statistical summary of our prior evolutionary experience and form the blueprint for the cognitive capacity of human brain. I will review the mathematical formalism that can explain a wide range of data on the interplay between molecular and experience-dependent mechanisms of connectivity formation. [Preview Abstract] |
Saturday, December 4, 2021 12:27PM - 12:39PM |
D04.00003: A mathematical metaplasticity model to stabilize spike timing-dependent plasticity. Zeyuan Wang, Luis Cruz In the brain, neurons are connected through neural networks that propagate signals in the form of voltage spikes. An important open question in biophysics is how the brain learns. Spike timing-dependent plasticity is a learning rule that is widely found in the brain. This rule is crucial for understanding how the brain learns and building efficient and functional artificial spiking neural networks. However, recent computational studies have shown that it has intrinsic run-away dynamics, causing the strengths of the connections to diverge. Here we aim to study how to stabilize it in a single neuron. We build a mathematical model based on a biological mechanism called metaplasticity, which modifies the speed and conditions of learning rules. This model includes negative feedback dynamics to balance the potentiation and depression of the strengths of connections. With this model, a single neuron fed with repeated input patterns in random order can achieve stable strengths of connections and produce stable output patterns. Our results indicate that spike timing-dependent plasticity can be stabilized with the mathematical metaplasticity model. Future research will address the behavior of this stabilized spiking timing-dependent plasticity in the context of a neural network. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700