Scholarlink

<p>The use of artificial intelligence-based algorithms in addressing engineering challenges has become increasingly prevalent. Machine learning plays a vital role in efficiently processing and analysing large datasets, speeding up the development of new compounds. This research offers a thorough and insightful examination to predict optical absorption properties of spinels, based on their composition, structure, and physical parameters. This research presents a hybrid machine learning framework combining quantum Wasserstein generative adversarial networks (QWGAN) and Bidirectional Long Short-Term Memory (Bi-LSTM) for predicting the optical absorption properties of spinel oxides (AB<sub>2</sub>O<sub>4</sub>). A dataset of 4548 spinel compounds was analysed using t-distributed Stochastic Neighbour Embedding (t-SNE) for feature reduction, followed by K-means clustering for composition grouping. Synthetic data generation via QWGAN addressed experimental data limitations, while Bi-LSTM models classified compounds with high optical stability. The model achieved strong predictive performance, with root mean square error (RMSE) and mean absolute error (MAE) of 10.5&#xa0;meV/atom and 0.518&#xa0;meV/atom, respectively. These results demonstrate the effectiveness of the proposed hybrid approach in predicting optical behaviour and identifying stable spinel materials for optical applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid ML model of quantum-embedded Bi-LSTM classifier for the prediction of stable spinel oxides

  • 期刊论文

  • 发表日期：

    2025年11月27日

查看全文
  • Rupam Bhaduri,
  • S. Manasa,
  • Shanmugasundaram Sakthivel,
  • Mani Karthik

摘要

The use of artificial intelligence-based algorithms in addressing engineering challenges has become increasingly prevalent. Machine learning plays a vital role in efficiently processing and analysing large datasets, speeding up the development of new compounds. This research offers a thorough and insightful examination to predict optical absorption properties of spinels, based on their composition, structure, and physical parameters. This research presents a hybrid machine learning framework combining quantum Wasserstein generative adversarial networks (QWGAN) and Bidirectional Long Short-Term Memory (Bi-LSTM) for predicting the optical absorption properties of spinel oxides (AB2O4). A dataset of 4548 spinel compounds was analysed using t-distributed Stochastic Neighbour Embedding (t-SNE) for feature reduction, followed by K-means clustering for composition grouping. Synthetic data generation via QWGAN addressed experimental data limitations, while Bi-LSTM models classified compounds with high optical stability. The model achieved strong predictive performance, with root mean square error (RMSE) and mean absolute error (MAE) of 10.5 meV/atom and 0.518 meV/atom, respectively. These results demonstrate the effectiveness of the proposed hybrid approach in predicting optical behaviour and identifying stable spinel materials for optical applications.