Quantum Circuit Pruning Achieves 47.7% Fidelity Gain Via Smart Approximation

Why It Matters

The technique dramatically boosts NISQ reliability, enabling deeper quantum algorithms without additional hardware upgrades. It offers a practical, software‑level path to extend the usable depth of near‑term quantum processors.

Key Takeaways

• Routing-aware pruning cuts two‑qubit gates up to 48.6%.
• Fidelity gains reach 47.7% on NISQ circuits.
• Targets small‑angle controlled rotations with high routing cost.
• Integrates directly into compilation flow, outperforming pre‑routing methods.
• Benefits variational quantum algorithms and quantum machine learning.

Pulse Analysis

Near‑term quantum computers, often labeled NISQ devices, are constrained by high error rates and limited qubit connectivity. Each logical interaction that spans distant qubits typically requires a chain of SWAP gates, compounding noise and eroding the overall fidelity of the computation. As quantum algorithms grow in size, especially variational quantum eigensolvers and quantum machine learning models, the routing overhead can dominate the error budget, making efficient compilation strategies essential for practical deployments.

The newly proposed pruning method tackles this challenge by quantifying the trade‑off between executing a small‑angle rotation and the fidelity loss incurred from the SWAPs needed to place the qubits correctly. Using a fidelity‑loss model that incorporates gate error rates and physical qubit distances, the compiler decides to omit a gate when its expected contribution to the final state is smaller than the noise introduced by routing. Simulated benchmarks on realistic grid architectures demonstrate up to a 48.6% reduction in two‑qubit gate count and a 47.7% boost in final‑state fidelity, outperforming traditional pruning that ignores routing costs.

Beyond the immediate performance gains, this routing‑aware approximation opens a pathway for software‑centric quantum error mitigation. By integrating directly into the compilation pipeline, developers can automatically adapt circuits to the topology of any NISQ device without manual tuning. Future work aims to extend the approach to non‑parametric gates, validate results on physical quantum hardware, and combine it with advanced routing algorithms. If widely adopted, such techniques could extend the effective depth of quantum programs, accelerating the transition from experimental prototypes to commercially viable quantum applications.