Circuit Knitting Optimises Variational Quantum Algorithms with Reduced Sampling Overhead

The limited number of qubits available in current quantum computers presents a major obstacle to running complex calculations, but researchers are exploring ways to overcome this challenge. Jun Wu, Jiaqi Yang, and Jicun Li, along with colleagues at IEEE, introduce a new framework called CKVQA that tackles this problem by intelligently dividing large quantum circuits into smaller, manageable pieces. This approach, known as circuit knitting, typically requires extensive computation, however, CKVQA incorporates a circuit architecture search to minimise this overhead and identify circuits that balance performance with efficiency. By optimising these smaller circuit components, the team demonstrates a significant reduction in computational demands while maintaining accuracy when applied to important quantum algorithms such as approximate optimisation and eigensolvers, paving the way for more practical quantum computation.

Research into variational quantum algorithms often encounters high sampling overhead, which increases exponentially with the complexity of the problem. This research introduces CKVQA, a framework that applies circuit knitting to variational quantum algorithms. By employing a quantum circuit architecture search adapted to this scenario, CKVQA aims to minimize this overhead by identifying parameterized quantum circuits that achieve a favorable balance between algorithmic performance and computational cost. Additionally, the researchers have developed a subcircuit-level optimization method to accelerate the training of variational quantum algorithms and reduce overall execution time.

Circuit Reduction and Modular Execution

Current quantum computers are limited in the number of qubits available, hindering progress on many promising algorithms. Researchers are exploring techniques to overcome this limitation by breaking down large quantum circuits into smaller, manageable pieces. Circuit cutting involves splitting a circuit and re-running disconnected parts, while circuit knitting goes further by reusing qubits to stitch together these fragments. A key challenge is selecting optimal cut points to minimize overhead associated with communication and measurement. Scaling these techniques to larger, more complex circuits remains a significant concern.

Variational quantum algorithms are a leading approach for near-term quantum computing, but they face challenges such as barren plateaus, where optimization becomes difficult. Researchers are developing adaptive algorithms and exploring the use of deep learning to optimize quantum circuit structure. Hardware-efficient algorithms and techniques like circuit cutting are also being investigated to mitigate noise and reduce circuit depth, improving trainability and overcoming the limitations of current quantum hardware.

Circuit Knitting Expands Quantum Algorithm Scope

Researchers have developed a new framework, called CKVQA, to address the limitations of current quantum hardware when running complex algorithms. CKVQA employs circuit knitting, a technique that breaks down large quantum circuits into smaller pieces that can be executed on existing devices, expanding the scope of what is computationally possible. The core innovation lies in minimizing a key drawback of circuit knitting: a substantial increase in the number of measurements required, known as sampling overhead. Traditional circuit knitting methods often suffer from an exponential rise in sampling overhead, negating the benefits of circuit decomposition.

CKVQA overcomes this by intelligently designing the structure of the quantum circuits themselves, tailoring them to work efficiently with the circuit knitting process. This co-optimization of circuit design and decomposition dramatically reduces the required measurements, making the approach practical for real-world applications. The framework utilizes a technique inspired by quantum architecture search, automatically identifying circuit structures that balance algorithmic performance and minimized sampling overhead. Results demonstrate a substantial reduction in sampling overhead while maintaining accuracy comparable to conventional circuit designs.

Furthermore, CKVQA incorporates a subcircuit-level optimization method that accelerates the training of variational quantum algorithms, a class of algorithms particularly well-suited for near-term quantum computers. By updating parameters within these smaller subcircuits, the framework enhances overall computational efficiency, demonstrating the potential to unlock more complex calculations and accelerate progress in fields like materials science and drug discovery. The framework represents a significant step towards realizing the full potential of near-term quantum computers by overcoming a key limitation in scalability.

Circuit Knitting and Quantum Architecture Search

This research introduces CKVQA, a framework designed to improve the efficiency of variational quantum algorithms on near-term quantum hardware. The core of CKVQA lies in its ability to combine circuit knitting, a technique for partitioning large quantum circuits, with a specialized quantum architecture search. This search automatically identifies circuit designs that minimize the computational overhead associated with circuit knitting while maintaining algorithmic performance. Furthermore, the framework incorporates a subcircuit optimization strategy that accelerates training and reduces overall execution time.

Results demonstrate that CKVQA effectively reduces sampling overhead and execution time without significantly impacting the accuracy of the algorithms tested. Notably, the quantum architecture search component can be performed entirely on classical hardware, decoupling the process from the need for quantum device access, a significant advantage given the limited availability of quantum resources. The authors acknowledge that future work could explore integrating CKVQA with other techniques to further enhance scalability and performance, representing a step towards bridging the gap between theoretical quantum algorithms and practical implementation on current quantum hardware.