Distributed Variational Quantum Algorithm Achieves 1,000-Variable Combinatorial Optimisation Solutions

Key Takeaways

• DVQA uses T-HOSVD to decompose quantum tensors.
• Classical tensor C replaces entanglement, reducing hardware demands.
• Scales to 1,000-variable MaxCut and portfolio optimisation.
• Noise impact limited to subsystem size, enhancing robustness.
• Validated on Wu Kong quantum computer hardware.

Pulse Analysis

Combinatorial optimisation has long been a stumbling block for noisy intermediate‑scale quantum (NISQ) processors, where the sheer number of variables quickly outpaces available qubits. DVQA tackles this bottleneck by applying truncated higher‑order singular value decomposition to the problem’s tensor representation, yielding a compact set of subsystem states. This tensor‑network strategy retains essential inter‑variable dependencies that traditional decomposition methods often discard, ensuring the algorithm’s expressive power remains intact while fitting within hardware constraints.

The core innovation lies in replacing costly quantum entanglement with a classical amplitude tensor, C, that encodes global correlations. Each subsystem runs on an isolated NISQ device, and local measurement results are stitched together through tensor contractions to form a distributed objective function. This architecture localises noise, making error rates proportional to subsystem size rather than the total qubit count. Empirical results show DVQA achieving near‑optimal solutions for MaxCut and portfolio optimisation problems involving up to 1,000 variables, and a successful hardware demonstration on the Wu Kong quantum computer underscores its practical viability.

Beyond immediate performance gains, DVQA reshapes the roadmap toward quantum advantage in optimisation. By demonstrating that large‑scale problems can be partitioned without sacrificing solution quality, the method opens doors for hybrid quantum‑classical pipelines in sectors such as finance, logistics, and energy. Future research aimed at extending the theoretical bounds to higher‑order Hamiltonians could further broaden its applicability, positioning DVQA as a cornerstone technology for the next generation of quantum‑enabled enterprise analytics.