Oscillator Ising Machines
ORAL · Invited
Abstract
For many real-world applications, modern society has become increasingly reliant on rapid and routine solution of hard discrete optimization problems.
Over the past decade, fascinating analog hardware approaches have arisen that combine principles of physics and computer science with optical, electronic and quantum engineering to solve combinatorial optimization problems in new ways---these have come to be known as Ising machines.
Such approaches leverage analog dynamics and physics to find good solutions of discrete optimization problems, potentially with advantages over traditional algorithms.
These approaches are based on the Ising model, a simple but powerful graph formulation with deep historical roots in the physics of magnetism, using which combinatorial optimization problems can be represented.
While the first Ising machines relied on quantum mechanical interactions, newer schemes have emerged that are based on purely classical (non-quantum) operational mechanisms.
Classical Ising machine schemes that can be implemented on chip have many practical advantages---eg., smaller physical size, lower cost, lower energy consumption, on-chip integration, scaling to large problem sizes and mass production.
About seven years ago, we discovered that the analog dynamics of networks of electronic oscillators could be exploited to solve Ising problems “naturally”.
In this talk, we will summarize the principles and practice of oscillator Ising machines, covering theory, computational characterization and fabricated hardware.
A key focus will be Ising machine performance on real-world applications --- for example, the MU-MIMO detection problem in modern wireless communications.
Our results indicate that near-optimal symbol-error rates (SERs) can be obtained, improving over the industrial state of the art by 20x for some scenarios.
Over the past decade, fascinating analog hardware approaches have arisen that combine principles of physics and computer science with optical, electronic and quantum engineering to solve combinatorial optimization problems in new ways---these have come to be known as Ising machines.
Such approaches leverage analog dynamics and physics to find good solutions of discrete optimization problems, potentially with advantages over traditional algorithms.
These approaches are based on the Ising model, a simple but powerful graph formulation with deep historical roots in the physics of magnetism, using which combinatorial optimization problems can be represented.
While the first Ising machines relied on quantum mechanical interactions, newer schemes have emerged that are based on purely classical (non-quantum) operational mechanisms.
Classical Ising machine schemes that can be implemented on chip have many practical advantages---eg., smaller physical size, lower cost, lower energy consumption, on-chip integration, scaling to large problem sizes and mass production.
About seven years ago, we discovered that the analog dynamics of networks of electronic oscillators could be exploited to solve Ising problems “naturally”.
In this talk, we will summarize the principles and practice of oscillator Ising machines, covering theory, computational characterization and fabricated hardware.
A key focus will be Ising machine performance on real-world applications --- for example, the MU-MIMO detection problem in modern wireless communications.
Our results indicate that near-optimal symbol-error rates (SERs) can be obtained, improving over the industrial state of the art by 20x for some scenarios.
*Support from the National Science Foundation, the Bakar Foundation and DARPA is gratefully acknowledged.
–
Presenters
-
Jaijeet Roychowdhury
- UC Berkeley