image: Qjump utilizes shallow quantum circuits to navigate the complex energy landscape and leap between promising regions—or basins—that often entrap traditional classical methods.
Credit: ©Science China Press
Quantum computing holds the promise of solving complex combinatorial optimization problems by encoding information into the low-energy states of Ising Hamiltonians. However, efficiently navigating these energy landscapes to find optimal solutions remains a formidable challenge for practical applications. To bridge this gap, a research team from Zhejiang University and their collaborators have developed "Qjump" (Quantum-enhanced jumping), a hybrid quantum-classical algorithm. Experimentally demonstrated on a 104-qubit superconducting processor, Qjump employs a strategic quantum sampling technique to accelerate the search for low-energy solutions, offering a new pathway toward practical quantum advantage.
Qjump employs shallow quantum circuits to navigate the complex "energy landscape" of a problem, enabling the system to leap between promising regions—or basins—that often entrap traditional classical methods. Once the quantum sampling identifies a candidate zone, a classical local search subroutine takes over to refine the solution. This hybrid strategy preserves high-quality results while drastically lowering the requirements for quantum circuit depth, providing a practical blueprint for achieving speedup on current hardware.
Guided by an analysis of quantum circuit dynamics, the team introduced a circuit truncation technique that streamlines the standard QAOA (Quantum Approximate Optimization Algorithm). This approach drastically reduces circuit depth and minimizes the impact of hardware noise, enabling the algorithm to "escape" suboptimal local minima more effectively. By integrating this streamlined quantum process with a classical local search, the researchers overcame the common trainability and noise-sensitivity issues that often cripple original QAOA performance, steering the search toward the most promising regions of the solution space.
To validate the algorithm, the researchers deployed Qjump on a 104-qubit superconducting processor to solve complex Ising problems. The experimental results demonstrated that Qjump consistently delivers higher-quality solutions than both the fixed-parameter QAOA and a highly optimized classical Simulated Annealing (SA) algorithm. Further efficiency analysis using the "time-to-solution" (TTS) metric reveals that on a near-term superconducting platform, Qjump is expected to achieve a 2.34-fold speedup for 104 qubits over single-core simulated annealing. While the team notes that this comparison focuses on sequential classical processing, the results provide a compelling benchmark for quantum-enhanced optimization on large-scale hardware.
Journal
National Science Review