BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Fully non-linear neuromorphic computing with linear wave scatterin
 g - Clara Wanjura\, Max Planck Institute for the Science of Light\, German
 y
DTSTART:20241111T143000Z
DTEND:20241111T150000Z
UID:TALK224050@talks.cam.ac.uk
CONTACT:Dr Sun-Woo Kim
DESCRIPTION:The increasing size of neural networks for deep learning appli
 cations and their energy consumption create a need for alternative neuromo
 rphic approaches\, for example\, using optics. Current proposals and imple
 mentations rely on physical nonlinearities or optoelectronic conversion to
  realise the required nonlinear activation function. However\, there are c
 onsiderable challenges with these approaches related to power levels\, con
 trol\, energy efficiency and delays. \n \nIn my talk\, I will present a sc
 heme [1] for a neuromorphic system that relies on linear wave scattering a
 nd yet achieves nonlinear processing with high expressivity. The key idea 
 is to encode the input in physical parameters that affect the scattering p
 rocesses. Moreover\, we show that gradients needed for training can be dir
 ectly measured in scattering experiments. We propose an implementation usi
 ng integrated photonics based on racetrack resonators\, which achieves hig
 h connectivity with a minimal number of waveguide crossings. Our work intr
 oduces an easily implementable approach to neuromorphic computing that can
  be widely applied in existing state-of-the-art scalable platforms\, such 
 as optics\, microwave and electrical circuits.\n \n[1] C.C. Wanjura\, F. M
 arquardt. Nat Phys 20\, 1434–1440 (2024).\n
LOCATION:Zoom link: https://zoom.us/j/92447982065?pwd=RkhaYkM5VTZPZ3pYSHpt
 UXlRSkppQT09
END:VEVENT
END:VCALENDAR