19 January 2024, at 4pm

 QAMSS (Quantum and Advanced Materials technologies for a Sustainable Society) seminar series, 4pm in the Cambridge Graphene Centre Seminar room, Electrical Engineering Division, 9 JJ Thomson Avenue.

Physics-Inspired Computing (π-computing) as an emergent paradigm of analogue computations. 


 Natalia G Berloff 

Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Wilberforce Road, Cambridge, CB3 0WA, United Kingdom


Pursuing superior computing speeds and enhanced power efficiency suggests going beyond the confines of conventional digital electronic systems. Analogue computing, guided by physics principles such as minimisation of energy, entropy or losses, is changing our approach to problem-solving on both hardware and software levels. Such physics-inspired computing systems function with continuous variables and present a viable strategy for tackling various challenges, from discrete combinatorial optimisation problems to machine learning tasks. This includes quadratic binary minimisation, a notable issue in the classical Ising minimisation of discrete spins, where both theoretical proposals and experimental realisations of physics-based analogue systems have been explored.

 In my talk, I will focus on emerging physical optimisers that utilise bifurcation dynamics and threshold operations to solve nonlinear problems. I will discuss various unconventional systems implementing optical neural networks and the problems suited to their architecture. In particular, I will discuss problems beyond spin Hamiltonians and show how to harness multiple degrees of freedom of physical optical systems for optimisation.