Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Photonics in Switching and Computing (PSC)

Not all Ising machines are created equal: choice of nonlinearity induces order-of-magnitude differences in computational performance

Fabian Böhm, Thomas Van Vaerenbergh, Guy Verschaffelt, Guy Van der Sande

Published Jan 1, 2021• 0 citations
Publisher

Abstract

Various nonlinear systems have been proposed for Ising machines, yet little insight exists into their suitability. We find the choice of nonlinearity to induce order-of-magnitude improvements in time-to-solution for saturable nonlinearities.
DOI
10.1364/psc.2021.m2b.4

Related papers

Order-of-magnitude differences in computational performance of analog Ising machines induced by the choice of nonlinearity

Fabian Böhm, Thomas Van Vaerenbergh, Guy Verschaffelt +1 more

Abstract Ising machines based on nonlinear analog systems are a promising method to accelerate computation of NP-hard optimization problems. Yet, their analog nature is also causing amplitude inhomogeneity which can deteriorate the ability to find optimal solutions. Here, we investigate how the system’s nonlinear transfer function can mitigate amplitude inhomogeneity and improve computational performance. By simulating Ising machines with polynomial, periodic, sigmoid and clipped transfer functions and benchmarking them with MaxCut optimization problems, we find the choice of transfer function to have a significant influence on the calculation time and solution quality. For periodic, sigmoid and clipped transfer functions, we report order-of-magnitude improvements in the time-to-solution compared to conventional polynomial models, which we link to the suppression of amplitude inhomogeneity induced by saturation of the transfer function. This provides insights into the suitability of nonlinear systems for building Ising machines and presents an efficient way for overcoming performance limitations.

arXiv:2012.10430•Communications Physics•Jul 1, 2021•63 citations

Chaotic-Amplitude-Controlled Coherent Ising Machines: Extensions to Quadratic Unconstrained Binary Optimisation and Simplified Models

Mastiyage Don Sudeera Hasaranga Gunathilaka, Sudeera Hasaranga Gunathilaka Mastiyage Don

•Mar 1, 2025•0 citations

Suppressing amplitude inhomogeneity enhances computational performance of Ising machines: why the choice of nonlinearity matters.

Fabian Böhm, Thomas Van Vaerenbergh, Guy Verschaffelt +1 more

Ising machines based on nonlinear analog systems are a promising method to accelerate computation of NP-hard optimization problems. Yet, their analog nature is also causing amplitude inhomogeneity which can deteriorate the ability to find optimal solutions. Here, we investigate if the shape of the system's nonlinear transfer function can mitigate amplitude inhomogeneity and improve computational performance. By simulating Ising machines with polynomial, periodic, sigmoid and clipped transfer functions and benchmarking them with MaxCut optimization problems, we find the choice of transfer function to have a significant influence on the calculation time and solution quality. For periodic, sigmoid and clipped transfer functions, we report order-ofmagnitude improvements in the time-to-solution compared to conventional polynomial models, which we link to the suppression of amplitude inhomogeneity induced by saturation of the transfer function. This provides insights into the suitability of systems for building Ising machines and presents an efficient way for overcoming performance limitations.

•Dec 18, 2020•2 citations

Chaotic-Amplitude-Controlled Coherent Ising Machines: Extensions to Quadratic Unconstrained Binary Optimisation and Simplified Models

Sudeera Hasaranga Gunathilaka Mastiyage Don

•Mar 1, 2025•0 citations

Combinatorial clustering with a coherent XY machine

Kyungduk Kim, Masahito Kumagai, Y. Yamamoto

We propose and numerically study the performance of an all-optical machine for tackling combinatorial clustering, one of the unsupervised machine learning problems. A problem instance is encoded into the phase of time-multiplexed optical pulses, which are coupled repulsively through optical delay lines. To maintain uniform pulse amplitudes, we utilize a nonlinear amplifier with gain saturation, enabling our optical architecture to emulate the classical XY-spin system. This solver, called the coherent XY machine, leads to the formation of clusters in an optical phase space and allows us to efficiently identify the solution with post-processing. Additionally, we implement momentum in our solver to provide a powerful mechanism for escaping local minima and searching for the global optimum. Benchmarking our approach with a most advanced Ising-spin-based solver reveals a two-orders-of-magnitude improvement in the time-to-solution of the algorithm. Furthermore, our approach exhibits scaling advantages for larger problems, which will facilitate time- and energy-efficient data clustering.

Optics Express•Aug 20, 2024•4 citations

How efficiently can we simulate the open system dynamics of Ising models

Anupam Mitra, Tameem Albash, Akimasa Miyake +1 more

Bulletin of the American Physical Society•Mar 16, 2021•0 citations
Abstract Ising machines based on nonlinear analog systems are a promising method to accelerate computation of NP-hard optimization problems. Yet, their analog nature is also causing amplitude inhomogeneity which can deteriorate the ability to find optimal solutions. Here, we investigate how the system’s nonlinear transfer function can mitigate amplitude inhomogeneity and improve computational performance. By simulating Ising machines with polynomial, periodic, sigmoid and clipped transfer functions and benchmarking them with MaxCut optimization problems, we find the choice of transfer function to have a significant influence on the calculation time and solution quality. For periodic, sigmoid and clipped transfer functions, we report order-of-magnitude improvements in the time-to-solution compared to conventional polynomial models, which we link to the suppression of amplitude inhomogeneity induced by saturation of the transfer function. This provides insights into the suitability of nonlinear systems for building Ising machines and presents an efficient way for overcoming performance limitations.
Ising machines based on nonlinear analog systems are a promising method to accelerate computation of NP-hard optimization problems. Yet, their analog nature is also causing amplitude inhomogeneity which can deteriorate the ability to find optimal solutions. Here, we investigate if the shape of the system's nonlinear transfer function can mitigate amplitude inhomogeneity and improve computational performance. By simulating Ising machines with polynomial, periodic, sigmoid and clipped transfer functions and benchmarking them with MaxCut optimization problems, we find the choice of transfer function to have a significant influence on the calculation time and solution quality. For periodic, sigmoid and clipped transfer functions, we report order-ofmagnitude improvements in the time-to-solution compared to conventional polynomial models, which we link to the suppression of amplitude inhomogeneity induced by saturation of the transfer function. This provides insights into the suitability of systems for building Ising machines and presents an efficient way for overcoming performance limitations.
We propose and numerically study the performance of an all-optical machine for tackling combinatorial clustering, one of the unsupervised machine learning problems. A problem instance is encoded into the phase of time-multiplexed optical pulses, which are coupled repulsively through optical delay lines. To maintain uniform pulse amplitudes, we utilize a nonlinear amplifier with gain saturation, enabling our optical architecture to emulate the classical XY-spin system. This solver, called the coherent XY machine, leads to the formation of clusters in an optical phase space and allows us to efficiently identify the solution with post-processing. Additionally, we implement momentum in our solver to provide a powerful mechanism for escaping local minima and searching for the global optimum. Benchmarking our approach with a most advanced Ising-spin-based solver reveals a two-orders-of-magnitude improvement in the time-to-solution of the algorithm. Furthermore, our approach exhibits scaling advantages for larger problems, which will facilitate time- and energy-efficient data clustering.