BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Training phase masks for diffractive networks / Emulating quantum 
 computing with optical matrix multiplication
DTSTART;VALUE=DATE-TIME:20241202T123000Z
DTEND;VALUE=DATE-TIME:20241202T125000Z
DTSTAMP;VALUE=DATE-TIME:20260906T122342Z
UID:indico-contribution-2473@events.chpc.ac.za
DESCRIPTION:Speakers: Hadrian Bezuidenhout (University of the Witwatersran
 d)\, Mwezi Koni (University of the Witwatersrand)\n**Training phase masks 
 for diffractive networks**\n\nDiffractive optical networks have been shown
  to be useful in a wide variety of application in the realms of optical co
 mputing and information processing\, such as modal sorting and multiplexin
 g. These systems utilise repeated phase modulations to transform a set of 
 input states contained within a particular basis into a set of target stat
 es contained in some new arbitrary basis. Traditionally\, these phase mask
 s are trained iteratively through randomly generated matrices or by optimi
 sing for individual pixels using various parameter search methods. In this
  work\, we construct a diffractive network capable of sorting various stru
 ctured light modes into pre-defined channels using optical aberrations. As
  such\, we leverage phase masks constructed from a superposition of Zernik
 e polynomials\, whose component weightings are trained analogously to the 
 weightings within a neural network. In this way\, we can solve for the des
 ired transformation by optimising for the weightings of fewer aberrations 
 rather than many individual pixels\, considerably reducing the number of v
 ariables needing to be solved.\n\n**Emulating quantum computing with optic
 al matrix multiplication**\n\nOptical computing offers a powerful platform
  for efficient vector-matrix operations\, leveraging the inherent properti
 es of light such as interference and superposition\, which also play a cru
 cial role in quantum computation. In this work\, we bridge classical struc
 tured light with quantum computing principles by reformulating photonic ma
 trix multiplication using quantum mechanical concepts like state superposi
 tion. We highlight the tensor product structure embedded in the cartesian 
 transverse degrees of freedom of light\, which serves as the foundation fo
 r optical vector-matrix multiplication. To demonstrate the versatility of 
 this approach\, we implement the Deutsch-Jozsa algorithm\, utilising a dis
 crete basis of localized Gaussian modes arranged in a lattice formation. A
 dditionally\, we illustrate the operation of a Hadamard gate by harnessing
  the programmability of spatial light modulators (SLMs) and Fourier transf
 orms through lenses. Our results underscore the adaptability of structured
  light for quantum information processing\, showcasing its potential in em
 ulating quantum algorithms.\n\nhttps://events.chpc.ac.za/event/139/contrib
 utions/2473/
LOCATION:Boardwalk Convention Centre BICC.G-D2 - D2 Tsitsikamma
URL:https://events.chpc.ac.za/event/139/contributions/2473/
END:VEVENT
END:VCALENDAR