Programming Quantum Hardware via Levenberg Marquardt Machine Learning

TL;DR

This paper introduces a machine learning approach using a modified Levenberg Marquardt method to program quantum hardware, demonstrating robustness to noise and decoherence, and successfully estimating quantum entanglement on IBM quantum devices.

Contribution

It presents a novel machine learning technique for quantum programming that reduces complexity and hardware requirements, applicable to noisy intermediate-scale quantum devices.

Findings

Successfully estimated quantum entanglement experimentally

Demonstrated robustness to noise and decoherence

Reduced hardware hits using the Levenberg Marquardt method

Abstract

Significant challenges remain with the development of macroscopic quantum computing, hardware problems of noise, decoherence, and scaling, software problems of error correction, and, most important, algorithm construction. Finding truly quantum algorithms is quite difficult, and many quantum algorithms, like Shor prime factoring or phase estimation, require extremely long circuit depth for any practical application, necessitating error correction. Machine learning can be used as a systematic method to nonalgorithmically program quantum computers. Quantum machine learning enables us to perform computations without breaking down an algorithm into its gate building blocks, eliminating that difficult step and potentially reducing unnecessary complexity. In addition, we have shown that our machine learning approach is robust to both noise and to decoherence, which is ideal for running on inherently noisy NISQ devices which are limited in the number of qubits available for error correction. We demonstrated this using a fundamentally non classical calculation, experimentally estimating the entanglement of an unknown quantum state. Results from this have been successfully ported to the IBM hardware and trained using a powerful hybrid reinforcement learning technique which is a modified Levenberg Marquardt LM method. The LM method is ideally suited to quantum machine learning as it only requires knowledge of the final measured output of the quantum computation, not intermediate quantum states which are generally not accessible. Since it processes all the learning data simultaneously, it also requires significantly fewer hits on the quantum hardware. Machine learning is demonstrated with results from simulations and runs on the IBM Qiskit hardware interface.