Quantum Circuit Distillation and Compression

TL;DR

This paper introduces a reinforcement learning-based method for distilling quantum circuits, producing shorter, noise-resistant circuits that maintain original functionality, thereby enhancing quantum computation on noisy hardware.

Contribution

The paper presents a novel quantum-circuit distillation technique using reinforcement learning, enabling the creation of shorter, noise-tolerant circuits for complex quantum algorithms.

Findings

Distilled circuits perform correctly on IBM-Quantum processors.

The method applies successfully to IQFT and Shor's algorithm.

A general rule for approximating n-qubit IQFTs was discovered.

Abstract

Quantum coherence in a qubit is vulnerable to environmental noise. When long quantum calculation is run on a quantum processor without error correction, the noise often causes fatal errors and messes up the calculation. Here, we propose quantum-circuit distillation to generate quantum circuits that are short but have enough functions to produce an output almost identical to that of the original circuits. The distilled circuits are less sensitive to the noise and can complete calculation before the quantum coherence is broken in the qubits. We created a quantum-circuit distillator by building a reinforcement learning model, and applied it to the inverse quantum Fourier transform (IQFT) and Shor's quantum prime factorization. The obtained distilled circuit allows correct calculation on IBM-Quantum processors. By working with the quantum-circuit distillator, we also found a general rule to generate quantum circuits approximating the general $n$-qubit IQFTs. The quantum-circuit distillator offers a new approach to improve performance of noisy quantum processors.