Hybrid Optimisation Improves Solutions for Large-Scale Ising Models Without Embedding

Quantum annealing holds promise for solving complex optimisation problems, but current machines struggle with large, intricate models, often requiring simplification that sacrifices solution quality. Shuta Kikuchi, from Keio University, Nozomu Togawa from Waseda University, and Shu Tanaka, also from Keio University, investigated a novel approach to improve performance by combining the strengths of classical and quantum computation. Their research introduces a hybrid optimisation method that uses a simulated annealing process to pre-process problems before submitting them to a quantum annealer, even when the original problem is too large to fit on the machine. The team’s findings demonstrate that this hybrid approach consistently yields better solutions than pre-processing alone, and crucially, reveals how the balance between pre-processing and quantum computation impacts overall accuracy, offering valuable insights for optimising future quantum annealing strategies

Hybrid Classical-Quantum Optimization for Ising Machines

Researchers are exploring ways to combine classical and quantum computing techniques to improve the performance of quantum annealing machines when solving complex problems. These machines, known as Ising machines, are designed to tackle challenging combinatorial optimization problems, which arise in many fields. Current quantum hardware faces limitations when dealing with large, complex problems, so this research focuses on hybrid methods that leverage the strengths of both classical and quantum approaches. The core idea is to use classical algorithms to simplify a problem before submitting it to a quantum annealer.

This pre-processing step reduces the complexity and size of the problem, making it more manageable for the quantum hardware. The research investigates techniques including fixing certain variables to reduce the number of possibilities, and iteratively alternating between classical and quantum steps to refine the solution. Understanding the behaviour of Ising machines is crucial to developing effective optimization strategies. Researchers are also investigating their statistical mechanics, analysing their dynamical properties, and exploring the role of chaotic behaviour and memory effects in their performance. This theoretical foundation informs the development of hybrid algorithms and helps to understand the limitations of current quantum annealing hardware.

Classical Pre-processing Boosts Quantum Annealing Performance

Researchers have developed a novel hybrid optimization method that improves the performance of quantum annealing machines when tackling complex problems. Recognizing that many real-world challenges exceed the capacity of current quantum hardware, the team combined a classical simulated annealing (SA) approach with a quantum annealing machine. This strategy leverages the strengths of both methods, using the classical algorithm to pre-process the problem and reduce its complexity before handing it over to the quantum hardware for refinement. The innovation lies in a two-stage process where the classical SA algorithm first “fixes” certain variables, effectively reducing the number of variables the quantum machine needs to consider.

This pre-processing creates a simplified “sub-Ising model” that is more manageable for the quantum annealer. By strategically fixing spins, the researchers aim to create a problem that can be more effectively embedded and solved by the quantum hardware, even if the original, larger problem is beyond its capacity. Extensive tests on large-scale Ising models demonstrate that solving the simplified sub-Ising model with the quantum annealer yields more accurate results than attempting to solve the original, complex problem directly. This suggests that the pre-processing step effectively reshapes the problem landscape, making it easier for the quantum machine to find optimal or near-optimal solutions. The number of fixed spins and the precision of the quantum annealer significantly influence the final solution quality and its dependence on the size of the simplified problem.

Hybrid Approach Boosts Optimization Problem Solving

Researchers are continually seeking ways to improve the performance of machines tackling complex optimization problems, which involve finding the best combination from a vast number of possibilities. These problems are common across many fields, from logistics and finance to scientific modelling. Ising machines, designed to solve these challenges, come in two main types: quantum annealing machines and their classical, simulated-annealing counterparts. This research focuses on a hybrid method that combines the strengths of both approaches, using a simulated annealing machine to preprocess a problem before passing it to a quantum annealer.

The team discovered that this combination consistently yields better solutions than using the simulated annealing machine alone, even when the problem is too large to fit directly onto the quantum annealer. A key finding is that solving a simplified version of the problem, created by reducing the number of variables, on the quantum annealer is more effective than attempting to solve the full, original problem. The effectiveness of this hybrid approach hinges on carefully balancing the number of variables fixed during preprocessing and the precision of the quantum annealer itself. The research demonstrates that a greater number of fixed variables, combined with a more accurate quantum annealer, leads to improved solution accuracy, even as the size of the remaining problem changes. This suggests a pathway to overcoming the limitations of both machine types, allowing researchers to tackle increasingly complex optimization challenges with greater efficiency and precision.

Hybrid Quantum Annealing Improves Problem Solving

This research presents a hybrid optimization method that combines a conventional Ising machine with a quantum annealing machine to improve solutions for complex problems. The team demonstrated that this approach yields better results than simply using the conventional machine alone, even when the problem is too large to be directly processed by the quantum annealer. The method involves strategically fixing certain spin values based on initial solutions obtained from the conventional machine, effectively creating a smaller, more manageable sub-problem for the quantum annealer to solve. Analysis reveals that the success of this hybrid method hinges on several factors, including the number of fixed spins and the accuracy of the quantum annealing machine itself.

Importantly, solving the reduced sub-problem with the quantum annealer proves more effective than attempting to solve the original, larger problem directly. The researchers found that the method’s performance is linked to the minimum energy gap of the sub-Ising model, suggesting that a carefully constructed sub-problem facilitates more accurate solutions. The authors acknowledge that the performance is also dependent on the size of the sub-Ising model and the characteristics of the initial solutions obtained from the conventional machine. Future work could explore strategies for optimizing the selection of fixed spins and further refining the balance between the conventional and quantum components of the hybrid approach.