A key challenge hinders scaling up quantum algorithms for real-world problems. Fast Reverse Iterative Optimisation, or FRIO, overcomes limitations in existing methods by harnessing continuous reverse quantum annealing across a transition into a many-body localised phase. This allows more efficient exploration of complex problem spaces; specifically, it improves performance against established techniques when finding solutions to challenging spin glass models like the Sherrington-Kirkpatrick model.

Fast Reverse Iterative Optimisation (FRIO) is a new quantum algorithm designed to tackle complex problems modelled by disordered systems known as spin glasses. It achieves better results and requires fewer computational steps than previous approaches such as Quantum Approximate Optimisation Algorithm. Researchers at Phasecraft and University of Leeds have developed Fast Reverse Iterative Optimisation (FRIO), a new quantum algorithm for tackling these challenges.

These scenarios, often used as models for understanding how energy behaves within materials exhibiting ‘many-body localisation’, present vital hurdles for existing methods such as Quantum Approximate Optimisation Algorithm; FRIO demonstrably improves upon these by requiring fewer computational steps. The team’s approach uses “continuous reverse quantum annealing”, which can be imagined as carefully unbending a paperclip instead of snapping it straight, crossing into this many-body localised phase where interactions prevent even energy distribution, think of trying to mix oil and water. This process is then approximated using simpler steps, similar to breaking down a long car journey into smaller segments, to enable use with current computers.

FRIO achieves constant-time ground state discovery via digitized reverse quantum annealing

The Fast Reverse Iterative Optimisation algorithm (FRIO) achieved a performance improvement over existing methods; it consistently outperformed conventional Quantum Approximate Optimisation Algorithm benchmarks when finding ground states for spin-glass problems like the Sherrington-Kirkpatrick model. Researchers digitised their continuous reverse quantum annealing approach into circuits suitable for near-term devices by using many-body localisation, interactions preventing even energy distribution, and developed a strategy employing thermalizing dynamics to automatically optimise circuit parameters.

Exploiting many-body localisation and accelerating continuous reverse quantum annealing via control of Landau-Zener crossings during quenching drives this improvement; the process involves gradually changing parameters within calculations. Digitising the approach into circuits with parameters determined through thermalizing dynamics allows efficient exploration of local minima. Numerical tests against Quantum Approximate Optimisation Algorithm (QAOA) demonstrated FRIO almost always outperformed QAOA when seeking ground states of the Sherrington-Kirkpatrick model, but current results are limited to relatively small systems, extending performance to larger problems relevant for real-world applications remains a challenge.

Scaling Quantum Annealing Reveals Potential Algorithm Acceleration

While FRIO offers improved performance over QAOA on notoriously difficult spin glass problems like the Sherrington-Kirkpatrick model, its validation currently remains limited; results haven’t extended beyond systems containing just 12. This raises questions about whether these gains will hold as scientists attempt to scale up calculations towards sizes relevant for practical applications, where computational demands often balloon exponentially. Despite this limitation, the work provides valuable insight into optimising quantum algorithms and accelerating continuous annealing by understanding transitions between different quantum behaviours.

Fast Reverse Iterative Optimisation (FRIO), a new quantum algorithm, aims to improve upon existing methods such as Quantum Approximate Optimisation Algorithm when solving complex problems found in spin glass models, these models mimic challenges encountered in materials science and machine learning. The team accelerated continuous reverse quantum annealing by considering Landau-Zener crossing density during quenching; they slowly reversed cooling until reaching low energy states. By exploring transitions into many-body localisation, where interactions limit energy distribution, FRIO seeks an improved scaling of required annealing time compared to previous approaches that scaled with system size *N*.

Fast Reverse Iterative Optimization (FRIO) is a new algorithm that outperformed the Quantum Approximate Optimisation Algorithm at finding ground states for small instances of the Sherrington-Kirkpatrick spin glass model, systems tested contained up to 12. This improvement stems from accelerating continuous reverse quantum annealing by analyzing Landau-Zener crossings during quenching, which suggests a reduction in necessary computation time as problem size increases. The researchers demonstrated this approach could be implemented using circuits requiring only modest numbers of steps, making it potentially suitable for current quantum computers and providing insight into optimising quantum algorithms.

👉 More information
🗞 Non-adiabatic Optimisation via Localisation-Delocalisation Transitions
✍️ Ryan Smith, Sabrina Yue Wang, Filippo Maria Gambetta and Raul A. Santos (Affiliation: Phasecraft)
🧠 ArXiv: https://arxiv.org/abs/2610.30301

More like this