×
 Physics at Virginia
ABSTRACT:

Understanding the emergent computational properties of the brain is challenging as we lack a theory of computation in the brain. A good place to start is by studying the dynamics of brain-like, neuromorphic, systems. Neuromorphic systems have properties akin to neurons; understanding the dynamical information processing properties of neuromorphic networks will aid in designing brain-like computers. Beyond applications, these are interesting systems due to their emergent properties, including phase transitions and near critical dynamics.

In this talk I will describe some neuromorphic systems and material properties necessary for a system to be considered neuromorphic, this will include discussing specific circuit implementations. Extending from nonequilibrium thermodynamics, I discuss a general formalism to understand circuits with memory. Using conservation laws we have developed a unifying formalism that relates local conservation laws to network topology. I will present recent work applying these techniques to train experimental neuromorphic hardware. The stable points of driven mem-circuits can be analyzed to provide insight into information embedded in a trained network and I will discuss a relationship to reproducing kernels. Finally, I will compare physical and artificial neural networks including how they learn different eigenstates.

 

Bio:

Frank Barrows is Director's Postdoctoral Fellow in The Physics of Condensed Matter and Complex Systems group at Los Alamos National Laboratory and a Center of Nonlinear Studies Fellow. He received his PhD in 2021 from Northwestern University, completing his graduate work at Argonne National Laboratory studying the emergent properties of neuromorphic and magnetic systems. He received his MD in 2023 from Northwestern University as part of the Medical Scientist Training Program. He graduated from the University of Virginia in 2013, majoring in Biology and Physics. He is interested in emergent computation in unconventional computing systems, neuromorphic systems and algorithms, nonequilibrium classical and quantum systems, noncommutative geometry, and consciousness.

Condensed Matter Seminar
Thursday, November 21, 2024
3:30 PM
Physics, Room 338

 Add to your calendar

To add a speaker, send an email to phys-speakers@Virginia.EDU. Please include the seminar type (e.g. Condensed Matter Seminars), date, name of the speaker, title of talk, and an abstract (if available).