Digital Zero-Noise Extrapolation with Quantum Circuit Unoptimization

2025-03-08

AI Summary

Zero-noise extrapolation (ZNE), a key quantum error mitigation technique, suffers from compiler-induced circuit simplification and insufficient noise-scaling diversity, undermining its robustness and accuracy. Method: This paper proposes a novel digital ZNE paradigm based on quantum circuit anti-optimization. It systematically constructs functionally equivalent circuit variants with significantly increased gate counts and highly heterogeneous structures, enabling controlled, diverse, and compiler-resilient noise amplification. Contribution/Results: To our knowledge, this is the first application of circuit anti-optimization to ZNE; it enables exponential generation of non-redundant circuits, inherently evading compiler simplification and enhancing feasibility for server-side deployment. In high-noise simulations of quantum volume and QAOA-based Max-Cut tasks, the method successfully recovers observable expectation values, demonstrating effective suppression of bias-error propagation and precise control over noise scaling.

Technology Category

• Machine Learning: Quantum Machine Learning
• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Security and Privacy: Large-scale security measurements

Abstract

Quantum circuit unoptimization is an algorithm that transforms a quantum circuit into a different circuit that uses more gate operations while maintaining the same unitary transformation. We demonstrate that this method can implement digital zero-noise extrapolation (ZNE), a quantum error mitigation technique. By employing quantum circuit unoptimization as a form of circuit folding, noise can be systematically amplified. The key advantages of this approach are twofold. First, its ability to generate an exponentially increasing number of distinct circuit variants as the noise level is amplified, which allows noise averaging over many circuit instances with slightly different circuit structure which mitigates the effect of biased error propagation because of the significantly altered circuit structure from quantum circuit unoptimization, or highly biased local noise on a quantum processor. Second, quantum circuit unoptimization by design resists circuit simplification back to the original unmodified circuit, making it plausible to use ZNE in contexts where circuit compiler optimization is applied server-side. We evaluate the effectiveness of quantum circuit unoptimization as a noise-scaling method for ZNE in two test cases using depolarizing noise numerical simulations: random quantum volume circuits, where the observable is the heavy output probability, and QAOA circuits for the (unweighted) maximum cut problem on random 3-regular graphs, where the observable is the cut value. We show that using quantum circuit unoptimization to perform ZNE can approximately recover signal from noisy quantum simulations.

Problem

• Implements digital zero-noise extrapolation using quantum circuit unoptimization.
• Amplifies noise systematically to mitigate biased error propagation.
• Recovers signal from noisy quantum simulations in test cases.

Innovation

• Quantum circuit unoptimization amplifies noise systematically
• Exponentially increases distinct circuit variants for noise averaging
• Resists circuit simplification, enabling ZNE with compiler optimization