QuantumNAS: Noise-Adaptive Search for Robust Quantum Circuits (2107.10845v5)

Abstract: Quantum noise is the key challenge in Noisy Intermediate-Scale Quantum (NISQ) computers. Previous work for mitigating noise has primarily focused on gate-level or pulse-level noise-adaptive compilation. However, limited research efforts have explored a higher level of optimization by making the quantum circuits themselves resilient to noise. We propose QuantumNAS, a comprehensive framework for noise-adaptive co-search of the variational circuit and qubit mapping. Variational quantum circuits are a promising approach for constructing QML and quantum simulation. However, finding the best variational circuit and its optimal parameters is challenging due to the large design space and parameter training cost. We propose to decouple the circuit search and parameter training by introducing a novel SuperCircuit. The SuperCircuit is constructed with multiple layers of pre-defined parameterized gates and trained by iteratively sampling and updating the parameter subsets (SubCircuits) of it. It provides an accurate estimation of SubCircuits performance trained from scratch. Then we perform an evolutionary co-search of SubCircuit and its qubit mapping. The SubCircuit performance is estimated with parameters inherited from SuperCircuit and simulated with real device noise models. Finally, we perform iterative gate pruning and finetuning to remove redundant gates. Extensively evaluated with 12 QML and VQE benchmarks on 14 quantum computers, QuantumNAS significantly outperforms baselines. For QML, QuantumNAS is the first to demonstrate over 95% 2-class, 85% 4-class, and 32% 10-class classification accuracy on real QC. It also achieves the lowest eigenvalue for VQE tasks on H2, H2O, LiH, CH4, BeH2 compared with UCCSD. We also open-source TorchQuantum (https://github.com/mit-han-lab/torchquantum) for fast training of parameterized quantum circuits to facilitate future research.

• The paper introduces a novel noise-adaptive co-search framework that optimizes both variational circuits and qubit mapping in NISQ devices.
• Leveraging a SuperCircuit-based search and evolutionary algorithm, the method decouples circuit design from parameter tuning to reduce computational costs.
• Iterative gate pruning enhances noise resilience, achieving up to 33% higher accuracy on QML tasks across 14 quantum devices.

Overview of QuantumNAS: Noise-Adaptive Search for Robust Quantum Circuits

QuantumNAS: Noise-Adaptive Search for Robust Quantum Circuits presents a novel framework focused on enhancing the noise resilience of quantum circuits in Noisy Intermediate-Scale Quantum (NISQ) computers. The research introduces an innovative noise-adaptive co-search strategy for variational circuits and qubit mappings, which are pivotal for tasks in quantum machine learning (QML) and variational quantum eigensolvers (VQE).

Key Contributions

The paper identifies quantum noise as a principal challenge in NISQ-era quantum computing. Traditional methods primarily mitigate noise at the gate or pulse level. QuantumNAS shifts focus to making the quantum circuits themselves inherently robust against noise.

1. SuperCircuit-Based Search: The framework introduces a novel concept of a SuperCircuit. This is constructed with multiple layers of parameterized gates and trained by sampling SubCircuits. This strategy allows for efficient exploration of a large design space, decoupling circuit design from parameter tuning, thereby avoiding the computationally expensive full training of all candidate circuits.
2. Evolutionary Co-Search: QuantumNAS employs an evolutionary search methodology to co-optimize circuit structures and their corresponding qubit mappings. The evolutionary algorithm benefits from a fast and reliable performance estimator, which utilizes classical simulations augmented with real device noise models.
3. Iterative Gate Pruning: After identifying a suitable SubCircuit, QuantumNAS applies fine-grained pruning to remove redundant gates. This process results in a slimmed-down circuit with comparable performance but reduced noise sources.

Experimental Evaluation

QuantumNAS is rigorously evaluated on 12 QML and VQE benchmarks across 14 quantum devices, significantly outperforming non-noise-aware methods and existing noise-adaptive schemes. For example, in QML tasks, QuantumNAS achieves over 95% accuracy in 2-class, 85% in 4-class, and 32% in 10-class classifications on real quantum computers.

1. Performance on Real Devices: The results demonstrate that QuantumNAS circuits exhibit superior noise resilience across diverse quantum machines by effectively mitigating quantum errors.
2. Efficiency and Scalability: The SuperCircuit approach allows for a scalable search process, reducing computational overhead by evaluating all potential SubCircuits through parameter inheritance.
3. Comparative Accuracy: Compared to conventional designs, QuantumNAS delayed the accuracy degradation caused by increased gate errors, facilitating higher-capacity models and achieving up to 33% higher accuracy.

Theoretical and Practical Implications

QuantumNAS sets a precedent in quantum computing by integrating noise-adaptive quantum circuit design with efficient co-search mechanisms. Theoretical implications of this work include the potential for constructing more robust quantum algorithms directly at the circuit level. Practically, QuantumNAS has immediate applications in enhancing the reliability of quantum computations on current NISQ devices, making it a valuable tool for both quantum algorithm developers and hardware engineers.

Future Directions

The research opens several prospective avenues:

• Quantum Feature Maps: Extension of noise-adaptive strategies to the design of quantum feature maps for machine learning.
• Barren Plateau Mitigation: Exploration of the potential for QuantumNAS to address training convergence issues associated with barren plateaus.
• Hybrid Ansatze: Further development of hybrid ansatze that are tailored for both hardware specifications and specific computational tasks.

Concluding Remarks

QuantumNAS provides a comprehensive methodology for generating robust quantum circuits, integrating advanced search techniques with practical noise considerations. The accompanying open-source TorchQuantum library supports rapid implementations and further innovations in the design and training of variational quantum circuits, marking a significant advance in the field of quantum computing.