Exploring Quantum Active Learning for Materials Design and Discovery

TL;DR

This paper investigates the integration of quantum algorithms into active learning frameworks for materials discovery, demonstrating potential improvements in search efficiency for certain datasets in materials science.

Contribution

It introduces quantum active learning (QAL) using quantum regressors and kernels, extending classical AL methods with quantum computing techniques for materials design.

Findings

QAL improved search efficiency in most cases

Performance correlated with data roughness

Potential for quantum methods in chemical space exploration

Abstract

The meeting of artificial intelligence (AI) and quantum computing is already a reality; quantum machine learning (QML) promises the design of better regression models. In this work, we extend our previous studies of materials discovery using classical active learning (AL), which showed remarkable economy of data, to explore the use of quantum algorithms within the AL framework (QAL) as implemented in the MLChem4D and QMLMaterials codes. The proposed QAL uses quantum support vector regressor (QSVR) or a quantum Gaussian process regressor (QGPR) with various quantum kernels and different feature maps. Data sets include perovskite properties (piezoelectric coefficient, band gap, energy storage) and the structure optimization of a doped nanoparticle (3Al@Si11) chosen to compare with classical AL results. Our results revealed that the QAL method improved the searches in most cases, but not all, seemingly correlated with the roughness of the data. QAL has the potential of finding optimum solutions, within chemical space, in materials science and elsewhere in chemistry.