[seminar] seminar Zdenka Kuncic

[Damjan Pelc]

Drage kolegice i kolege,
zadovoljstvo nam je pozvati vas na sljedeći seminar Fizičkog odsjeka:

Dear colleagues,
it is our pleasure to invite you to the following seminar of the Physics
Department:

Srijeda/Wednesday *6.7.2022. 14 h, F-201*

Phase transitions, criticality and brain-inspired learning in physical
neural networks

Zdenka Kuncic
University of Sydney

Abstract:
Biological neural networks, unlike artificial neural networks, exhibit
collective dynamics that emerge from the many-body interactions between
neurons. If we could strip away the biological processes, new clues
might be revealed pointing to the physical mechanisms underlying the
brain's efficient information processing capabilities and these
mechanisms could then be exploited to develop next-generation
intelligent computing systems. In this talk, I will present efforts
towards demonstrating this principle using inorganic nanowires that
self-assemble into a complex network structure, similar to the brain's
neural network circuitry. Under electrical stimulation, nanowire
cross-point junctions exhibit hysteresis and synapse-like responses
that, collectively, are strikingly similar to the nonlinear dynamics
observed in the brain. We found evidence of first-order dynamical phase
transitions as well as critical dynamics. By harnessing the diverse
dynamics in the readouts from the nanowire physical neural network, we
also found optimal performance was achieved in a machine learning task
when the network was in a critical-like state, as has also been proposed
for the brain's neural network. Overall, these results demonstrate the
importance of network dynamics in brain-inspired physical learning and
have broader implications for neuromorphic systems and artificial
intelligence.


References:
Hochstetter J. et al. Avalanches and edge-of-chaos learning in
neuromorphic nanowire networks. Nat. Commun. 12, 4008 (2021) |
doi.org/10.1038/s41467-021-24260-z
Kuncic Z. & Nakayama T. Neuromorphic nanowire networks: principles,
progress and future prospects for neuro-inspired information processing.
Adv. Phys. X 6, 1894234 (2021) | doi.org/10.1080/23746149.2021.1894234

Biography:
Zdenka Kuncic is a Professor of Physics at the School of Physics,
University of Sydney, Australia, where she completed a BSc with
1st-class Honours in Physics. She completed a PhD in 1996 in theoretical
astrophysics from the University of Cambridge, UK, after which she was
awarded the prestigious 1851 Royal Commission International Research
Fellowship in Science and Engineering (former recipients of which
include Nobel laureates Dirac, Rutherford and Higgs), which she held at
the University of Victoria, BC, Canada. After returning to Sydney in
2000 to take up an academic position at her alma mater, she pivoted her
research direction towards medical and biological physics, in which she
has made a number of contributions to biomedical imaging physics,
systems biology, nanotechnology in medicine, neurotechnology and
neuromorphic systems. She is a former Australia-Harvard Fellow and
Australian-American Fulbright Future Scholar, as well as an elected
Fellow of the Australian Institute of Physics.

srdačno/best regards
Sanjin Benić & Damjan Pelc
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.phy.pmf.unizg.hr/pipermail/seminar/attachments/20220630/db59c4e0/attachment-0001.htm>