Orbit-QAOA breakthrough slashes quantum computing training time by 82%

Researchers have developed a new quantum computing method called Orbit-QAOA to improve training efficiency in solving complex problems. The technique, created by a team including Frank Phillipson and Daniel Dijkzeul at the Centrum Wiskunde & Informatica (CWI) in Amsterdam, reduces training time while maintaining high accuracy. It addresses key challenges in quantum optimisation by refining how parameters are updated during calculations.

The standard Multi-angle Quantum Approximate Optimisation Algorithm (MA-QAOA) often struggles with balancing computational cost and performance. Experiments showed that parameter settings in shallow circuits rarely match those in deeper ones, making fixed training methods inefficient. Orbit-QAOA solves this by cyclically revisiting and selectively refining layers instead of processing them in a rigid sequence.

By optimising the granularity of parameter updates and tracking gradients more intelligently, the method cuts training steps by up to 81.8%. It also reduces approximation errors by as much as 72 times compared to traditional approaches. Despite these gains, the technique still delivers the same level of accuracy as standard MA-QAOA.

The approach not only speeds up training but also lowers computational costs. This makes it easier to handle larger and more complex problems on near-term quantum devices. Beyond QAOA, the framework could also help optimise other parameterised quantum circuits, potentially benefiting fields that rely on combinatorial optimisation.

Orbit-QAOA offers a more efficient way to train quantum algorithms without compromising accuracy. Its ability to reduce both time and computational demands may expand the range of problems solvable with current quantum hardware. The method's flexibility also suggests broader applications in optimising different types of quantum circuits.