Quantum neural network (QNN) models have received increasing attention owing to their strong expressibility and resistance to overfitting. It is particularly useful when the size of the training data is small, making it a good fit for materials informatics (MI) problems. However, there are only a few examples of the application of QNN to multivariate regression models, and little is known about how these models are constructed. This study aims to construct a QNN model to predict the melting points of metal oxides as an example of a multivariate regression task for the MI problem. Different architectures (encoding methods and entangler arrangements) are explored to create an effective QNN model. Shallow-depth ansatzs could achieve sufficient expressibility using sufficiently entangled circuits. The "linear" entangler was adequate for providing the necessary entanglement. The expressibility of the QNN model could be further improved by increasing the circuit width. The generalization performance could also be improved, outperforming the classical NN model. No overfitting was observed in the QNN models with a well-designed encoder. These findings suggest that QNN can be a useful tool for MI.
Quantum neural network (QNN) models have received increasing attention owing to their strong expressibility and resistance to overfitting. It is particularly useful when the size of the training data is small, making it a good fit for materials informatics (MI) problems. However, there are only a few examples of the application of QNN to multivariate regression models, and little is known about how these models are constructed. This study aims to construct a QNN model to predict the melting points of metal oxides as an example of a multivariate regression task for the MI problem. Different architectures (encoding methods and entangler arrangements) are explored to create an effective QNN model. Shallow-depth ansatzs could achieve sufficient expressibility using sufficiently entangled circuits. The "linear" entangler was adequate for providing the necessary entanglement. The expressibility of the QNN model could be further improved by increasing the circuit width. The generalization performance could also be improved, outperforming the classical NN model. No overfitting was observed in the QNN models with a well-designed encoder. These findings suggest that QNN can be a useful tool for MI.
Since the introduction of the quantum neural network model, it has been widely studied due to its strong expressive power and robustness to overfitting. To date, the model has been evaluated primarily in classification tasks, but its performance in practical multivariate regression problems has not been thoroughly examined. In this study, the Auto-MPG data set (392 valid data points, excluding missing data, on fuel efficiency for various vehicles) was used to construct QNN models and investigate the effect of the size of the training data on generalization performance. The results indicate that QNN is particularly effective when the size of training data is small, suggesting that it is especially suitable for small-data problems such as those encountered in Materials Informatics.
Quantum Neural Networks (QNNs) have gained significant attention due to their strong expressibility and potential learning. The effectiveness of QNNs is impacted by the choice of input embedding when training samples are limited. Numerous QNN architectures have been reported in the literature; however, the effects of input embedding and entanglement strategies have yet to be examined. In this study, a comprehensive investigation of different embedding techniques, such as θ = x . π / Δ and θ = a r c t a n ( x ) + π / Δ , has been used in a QNN to predict the melting point of the metal oxide, where the dataset is limited to only 70 samples. Further, multiple scaling factors ( Δ ) for each embedding with linear and circular entanglements, along with variational circuit depths, have been examined. The robustness of each embedding has been evaluated using 21 independent random parameter initialisations, while prediction performance has been statistically evaluated using R 2 and RMSE with a generalisation gap. From the results, it is observed that the embeddings such as θ = a r c t a n ( x ) + π / 4 and θ = a r c t a n ( x ) + π / 5 have been performed invariably across random initialisation and achieved superior results compared with other embeddings. Finally, the performance of the QNN has been compared with traditional machine learning algorithms, and the results show that it outperforms the others.
Quantum neural networks (QNNs) are a revolutionary paradigm placed at the frontier of machine learning and quantum computing. By utilizing quantum mechanical principles like superposition, entanglement, and quantum parallelism, QNNs facilitate the solution of complex machine learning problems that were thought to be impossible according to conventional models. This chapter provides a detailed study of QNNs, from elementary principles to practical applications. It begins with a rundown of the most significant concepts in machine learning and quantum computing, followed by QNN architecture, with important points including quantum gates and parameterized quantum circuits. The chapter addresses challenges in training QNNs, with special attention given to hybrid quantum-classical optimization algorithms and their utility in successful model training. It also examines multiple applications of QNNs, including simulation in quantum chemistry, optimization problems, and real-world machine learning applications, to demonstrate how they have the potential to outperform conventional algorithms in high dimensional or highly entangled data environments. This chapter provides a comparative assessment of quantum machine learning frameworks, quantum hardware platforms, and hybrid optimization methods and aids in determining the most suitable tools to design QNNs. It finally discusses possible areas of future research and research issues in the field, emphasizing the urgent need to improve quantum hardware, increase noise tolerance, and scale up. As quantum computing technology evolves, QNNs are expected to be at the center of solving some of the most complex and computationally intensive problems encountered in various industries.
In this article, we introduce a novel approach for enhancing the performance of Quantum Neural Networks (QNNs) through variations in entanglement schemes within variational forms. Our main objective in this research is to improve the performance of QNNs, specifically focusing on their application in classification tasks. We tested three models, each with a different type of entanglement scheme: Linear Entanglement, Circular Entanglement, and our proposed Full Entanglement. Each model consisted of three layers of Ry parameterized gates. We used the Breast Cancer Wisconsin (Diagnostic) dataset and encoded it using the ZZ-feature Map. Then we used this encoded data for training our models and evaluated them based on metrics such as training and validation loss. Our experimental results demonstrate that increasing entanglement can mitigate overfitting, with the proposed Full Entanglement scheme showing the most promising results.
Quantum neural networks (QNNs) -- parameterised quantum circuits trained to perform pattern recognition and predictive tasks -- represent a prominent but contested direction in quantum machine learning, offering potential parameter efficiency and expressibility benefits over classical neural networks for specific problem structures while suffering from barren plateau trainability limitations at scale. The 2024 issue of this journal established SNAD LSAD-structured ansatz (Silva et al., 2024) as the critical enabler of scalable QNN training, and QHDC (Bianchi et al., 2024) showed that QNN advantage is dataset-dependent. This paper advances the QNN field with the Quantum Neural Network Design and Evaluation (QNNDE) framework, a systematic methodology for QNN architecture design, training optimisation, and performance evaluation across four pattern recognition and prediction task classes: temporal sequence modelling (quantum recurrent circuits), spatial pattern recognition (quantum convolutional circuits), generative modelling (quantum generative adversarial networks), and time-series regression (quantum reservoir computing). QNNDE is evaluated on eight benchmark tasks spanning physics simulation, molecular property prediction, quantum state tomography, financial time-series, medical signal classification, and quantum circuit synthesis on IBM Quantum Eagle (127-qubit) and IonQ Forte (35-qubit). Quantum recurrent circuits achieve 12.4pp accuracy improvement over classical LSTMs for quantum state evolution prediction. Quantum convolutional circuits match classical CNNs on image-derived molecular patterns at 68.4% parameter reduction. Quantum GANs generate molecular conformations with 8.4% lower FID than classical GAN at equal network size. Quantum reservoir computing outperforms classical echo state networks by 18.4% on quantum-process time-series prediction.
ABSTRACT Quantum computing has become increasingly practical in solving real‐world problems due to advances in hardware and algorithms. In this paper, we aim to design, apply, and evaluate quantum machine learning and hybrid quantum‐classical models in a few practical materials science tasks, i.e., predicting stacking fault energies and solutes that can ductilize magnesium. To this end, we adopt two different representative quantum algorithms, i.e., quantum support vector machines (QSVM) and quantum neural networks (QNN), and adjust them to our application scenarios. We systematically test the performance with respect to the hyperparameters of selected ansatzes. Eventually, we construct quantum models with optimized parameters for regression and classification that predict targeted solutes based on the elemental volumes, electronegativities, and bulk moduli of chemical elements. We identify a few combinations of hyperparameters that yield validation scores of approximately 90% for QSVM and hybrid QNN in both tasks.