Noise-Aware Quantum Architecture Search Achieves Robust Circuits with Nsga-Ii Algorithm

Researchers are tackling the critical challenge of building robust quantum circuits, as noise remains a significant barrier to realising practical quantum computation. Chenlu Li, Hui Zeng, and Dazhi Ding from the School of Microelectronics at Nanjing University of Science and Technology, along with et al., present a novel noise-aware quantum architecture search (NA-QAS) framework grounded in circuit design principles. Their work distinguishes itself by directly incorporating a noise model into the training process, identifying architectures inherently resilient to errors. By combining a Hamiltonian-greedy strategy with an enhanced NSGA-II algorithm, the team efficiently navigates the complex design landscape, automatically balancing performance with hardware limitations , a crucial step towards scalable and reliable quantum technologies. This innovative approach demonstrably outperforms existing methods in noisy conditions, paving the way for more practical and efficient quantum devices.

Noise-robust Quantum Circuit Design via Automated Search

Scientists have developed a novel noise-aware quantum architecture search (NA-QAS) framework to automate the design of high-performance quantum circuits, addressing limitations in current variational quantum algorithms (VQAs). This breakthrough research focuses on identifying noise-robust architectures tailored to specific tasks and hardware constraints, a critical challenge for near-term quantum devices. The team achieved this by incorporating a noise model directly into the training process of parameterized quantum circuits (PQCs), enabling the discovery of circuits resilient to environmental errors like bit flips and decoherence. A key innovation lies in the introduction of a hybrid Hamiltonian ε-greedy strategy, designed to optimise evaluation costs and effectively avoid getting trapped in suboptimal solutions during the search process.
The study unveils an enhanced variable-depth NSGA-II algorithm, meticulously crafted to navigate the expansive search space of possible quantum architectures. This algorithm facilitates an automated trade-off between architectural expressibility, the circuit’s ability to represent complex functions, and the practical limitations of quantum hardware overhead, such as gate count and circuit depth. Researchers implemented a parameter-sharing strategy, coupling multiple classical linear layers within a ‘supernet’ to streamline the optimisation process and reduce computational demands. This allows for the selection of the optimal layer for back-propagation, while all candidate architectures share a common set of quantum circuit parameters, jointly optimised through the hybrid Hamiltonian approach.

Experiments demonstrate the effectiveness of the NA-QAS framework through both binary classification and iris multi-classification tasks performed under simulated noisy conditions. The results clearly show that the proposed framework can search for quantum architectures exhibiting superior performance and greater resource efficiency compared to existing methods in the presence of noise. This work establishes a significant advancement in quantum algorithm design, offering a pathway towards more robust and practical quantum computations. The research opens exciting possibilities for developing quantum machine learning algorithms that can function reliably on noisy intermediate-scale quantum (NISQ) devices.

Furthermore, the team’s approach incorporates simulations of realistic device noise, including bit-flip and decoherence channels, directly into the fitness function of the NSGA-II algorithm. This allows for simultaneous optimisation of quantum circuit performance, measured by the expectation value of the task Hamiltonian, alongside hardware costs. By integrating noise-aware quantum neural networks, the hybrid Hamiltonian parameter-sharing strategy, and multi-objective evolutionary search, the NA-QAS framework provides a comprehensive solution for automated quantum architecture design. The findings validate the framework’s ability to identify Pareto-optimal trade-offs between performance, noise robustness, and implementation complexity, paving the way for more efficient and reliable quantum algorithms.

Noise-aware quantum architecture search via supernet optimisation offers

Scientists have developed a noise-aware quantum architecture search (NA-QAS) framework to identify robust architectures for parameterized quantum circuits (PQCs). This work addresses critical limitations in current quantum architecture search methods, namely the computational burden of evaluation, the barren plateau phenomenon hindering gradient computation, and the substantial data exchange between classical and quantum systems. To overcome these challenges, the research team pioneered a hybrid Hamiltonian-greedy strategy that optimises evaluation costs and avoids becoming trapped in local optima during the search process. The core of NA-QAS lies in a parameter-sharing strategy coupling multiple classical linear layers into a ‘supernet’ for quantum architectures.

For each sampled ansatz, the optimal linear layer is selected for backpropagation, enabling joint optimisation of all candidate architectures with a common set of quantum circuit parameters, a technique that dramatically reduces computational overhead and prevents premature convergence. Furthermore, researchers engineered an enhanced variable-depth NSGA-II evolutionary algorithm to navigate the expansive search space, simultaneously optimising quantum circuit performance, CNOT gate count, and circuit depth. The algorithm’s fitness function incorporates simulations of practical device noise, including bit-flip and decoherence channels, ensuring the identification of Pareto-optimal trade-offs between performance, noise robustness, and implementation complexity. Experiments employed binary classification and iris multi-classification tasks under noisy conditions to validate the framework’s effectiveness.

The study’s innovative approach integrates noise-aware quantum neural networks, the hybrid Hamiltonian parameter-sharing strategy, and multi-objective evolutionary search into a single automated framework. Results demonstrate that NA-QAS consistently identifies architectures with superior performance and greater resource efficiency compared to existing approaches in noisy environments, achieving a significant breakthrough in practical quantum circuit design. This method enables the creation of quantum circuits that maintain high fidelity even with real-world imperfections.

Noise-Robust Quantum Architectures via Automated Search

Scientists have developed a noise-aware architecture search (NA-QAS) framework for designing high-performance circuits, addressing a critical need in quantum computing . The research introduces a method for identifying noise-robust architectures by incorporating a noise model directly into the training of parameterized quantum circuits (PQCs). Experiments revealed that this approach enables the automated identification of a Pareto-optimal trade-off between quantum architecture performance, noise robustness, and implementation complexity. The team measured the effectiveness of NA-QAS through binary classification and iris multi-classification tasks conducted under simulated noisy conditions.

Results demonstrate that the framework consistently identifies quantum architectures with superior performance and greater resource efficiency compared to existing approaches under these noisy conditions. A hybrid Hamiltonian-greedy strategy was implemented to optimise evaluation costs and effectively circumvent local optima during the search process. This innovative strategy significantly reduces computational demands while maintaining search quality, a key advancement for practical application. Furthermore, researchers employed an enhanced variable-depth NSGA-II algorithm to navigate the vast search space, facilitating an automated trade-off between architectural expressibility and hardware overhead.

Data shows that this algorithm efficiently explores the design landscape, identifying optimal architectures that balance performance with resource constraints. The study details how the framework integrates noise-aware quantum neural networks, the Hamiltonian parameter-sharing strategy, and multi-objective evolutionary search to achieve these results. Tests prove that the NA-QAS framework successfully addresses the challenge of exponentially growing search spaces in quantum circuit design. The work details how the framework can generate circuits with increased quantum circuit depth and a greater number of qubits, essential for tackling more complex computational problems. Measurements confirm that the proposed methodology offers a significant improvement in identifying architectures suitable for implementation on noisy intermediate-scale quantum (NISQ) devices, paving the way for more robust and reliable quantum computations.

Noise-Robust Quantum Architectures via Automated Search offer promising

Scientists have developed a noise-aware quantum architecture search (NA-QAS) framework to automate the design of high-performance circuits for specific tasks. This new framework incorporates a noise model into the training of parameterized quantum circuits, identifying architectures robust to noise, a significant challenge in quantum computing. Researchers introduced a hybrid Hamiltonian-greedy strategy to optimise evaluation costs and avoid getting stuck in suboptimal solutions, alongside an enhanced variable-depth NSGA-II algorithm to balance architectural complexity and hardware demands. The effectiveness of NA-QAS was demonstrated through binary classification and iris multi-classification tasks performed under noisy conditions.

Results indicate that the framework successfully searches for architectures exhibiting superior performance and greater resource efficiency compared to existing methods, achieving perfect accuracy in the Iris classification task with only four qubits. The identified architectures offer a practical balance between expressibility and quantum resource overhead, potentially enabling scalable variational quantum algorithm (VQA) deployment on near-term hardware. Authors acknowledge that the data supporting their findings are not currently publicly available due to technical and logistical constraints. Future research could focus on addressing these limitations and exploring the framework’s performance on a wider range of quantum algorithms and hardware platforms, potentially broadening its applicability and impact. This work was supported by the National Natural Science Foundation of China and the Key Project of Science Foundation of Jiangsu Province.