Quadratic Binary Optimization Techniques

Quadratic binary optimisation techniques address decision problems where binary variables interact through pairwise terms, forming a quadratic objective function. Such problems, often expressed as QUBO or equivalent Ising models, are NP-hard and underpin applications in logistics, finance, machine learning and materials discovery. Solution strategies span classical exact methods—such as branch-and-bound and cutting-plane algorithms—and a rich array of heuristics including tabu search, variable neighbourhood search and specialised local-search move operators. In parallel, quantum approaches exploit annealing and gate-based devices to traverse rugged energy landscapes. Transformations of higher-order pseudo-Boolean functions into quadratic form via auxiliary variables have broadened the class of tractable problems, while hardware-aware preprocessing and parameter compression techniques have improved robustness against limited precision and noise. This multifaceted field continues to evolve through interplay between theoretical advances and practical implementations, driving global collaboration across optimisation, computer science and quantum engineering.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Developments in 2023–2024 have spanned problem formulation, algorithmic efficiency and hardware-aware preprocessing. A novel feature-selection method recasts the task of choosing a fixed number of predictive variables as a QUBO, enabling direct embedding on both classical solvers and quantum annealers. Numerical experiments on benchmark data sets demonstrate superior model quality compared with greedy and iterative schemes. Concurrently, theoretical work on parameter compression uses bounds on the global minimum to minimise the dynamic range of QUBO coefficients without altering the set of optima; empirical tests on random instances and real-world encodings—such as clustering and subset-sum problems—show significant gains in annealer performance under finite precision. On the algorithmic front, a closed-form analysis of r-flip neighbourhoods has produced criteria to prune candidate moves when single-bit flips yield no improvement. Integrating this r-flip strategy into multi-start tabu search delivers rapid convergence to high-quality solutions on very large QUBO benchmarks, underscoring the impact of refined combinatorial insights on local-search efficiency.

Quadratic Binary Optimization Techniques publication trend

The graph below shows the total number of articles in quadratic binary optimization techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Quadratic Unconstrained Binary Optimisation (QUBO): An NP-hard problem in which a quadratic polynomial over binary variables is minimised or maximised without explicit constraints.

Quantum annealing: A heuristic quantum procedure that exploits quantum fluctuations to explore energy landscapes and identify low-energy states of Ising or QUBO models.

r-flip move: A local search operation that simultaneously flips the values of r binary variables to transition between neighbouring solutions.

Dynamic range: The span between the largest and smallest coefficients in a QUBO instance, influencing hardware precision requirements and solution stability.

References

1. Feature selection on quantum computers. Quantum Machine Intelligence (2023).
2. An Efficient Closed-Form Formula for Evaluating r-Flip Moves in Quadratic Unconstrained Binary Optimization. Algorithms (2023).
3. Optimum-preserving QUBO parameter compression. Quantum Machine Intelligence (2024).