Hybrid Quantum Graph Neural Network for Molecular Property Prediction

TL;DR

This work presents HyQCGNN, a gradient-free hybrid quantum-classical graph neural network designed to predict formation energies of oxide perovskites. It combines a classical GENConv layer with a quantum layer that uses amplitude encoding over 5 qubits and a trainable ansatz, optimized via Nevergrad NGOpt. On the MatBench perovskite dataset, HyQCGNN is competitive with classical baselines like CGNN and XGBoost, though it does not yet outperform them, highlighting both the promise and current limits of quantum feature encoding in graph-based materials prediction. The study demonstrates a workable architecture and workflow for integrating quantum circuits into graph neural networks, suggesting that advances in quantum embedding strategies and hardware could yield tangible benefits for complex materials ML tasks.

Abstract

To accelerate the process of materials design, materials science has increasingly used data driven techniques to extract information from collected data. Specially, machine learning (ML) algorithms, which span the ML discipline, have demonstrated ability to predict various properties of materials with the level of accuracy similar to explicit calculation of quantum mechanical theories, but with significantly reduced run time and computational resources. Within ML, graph neural networks have emerged as an important algorithm within the field of machine learning, since they are capable of predicting accurately a wide range of important physical, chemical and electronic properties due to their higher learning ability based on the graph representation of material and molecular descriptors through the aggregation of information embedded within the graph. In parallel with the development of state of the art classical machine learning applications, the fusion of quantum computing and machine learning have created a new paradigm where classical machine learning model can be augmented with quantum layers which are able to encode high dimensional data more efficiently. Leveraging the structure of existing algorithms, we developed a unique and novel gradient free hybrid quantum classical convoluted graph neural network (HyQCGNN) to predict formation energies of perovskite materials. The performance of our hybrid statistical model is competitive with the results obtained purely from a classical convoluted graph neural network, and other classical machine learning algorithms, such as XGBoost. Consequently, our study suggests a new pathway to explore how quantum feature encoding and parametric quantum circuits can yield drastic improvements of complex ML algorithm like graph neural network.

Figures (4)

• Figure 1: Featurization overview
• Figure 2: Quantum circuit implemented in our hybrid QGCNN model which includes a data loading layer using amplitude encoding, a variational layer, and a trainable readout layer. For space considerations the amplitude encoding layer is restricted to 3-qubits
• Figure 3: Plot and associated $R^2$ value for the true formation energy vs different models' prediction of formation energy.
• Figure 4: Feature importances reported by the XGBoost algorithm. The parenthesis indicate which atom or pair of atoms the feature applies to.