Matini-Net: Versatile Material Informatics Research Framework for Feature Engineering and Deep Neural Network Design DOI
Myeonghun Lee, Taehyun Park, Kyoungmin Min

et al.

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 21, 2024

In this study, we introduced Matini-Net, which is a versatile framework for feature engineering and automated architecture design materials informatics research using deep neural networks. Matini-Net provides the flexibility to feature-based, graph-based, combinations of these models, accommodating both single- multimodal model architectures. For validation, performed performance evaluation on MatBench benchmarking dataset five properties, targeting types regression architectures that can be designed Matini-Net. When applied each material property datasets, best various exhibited

Language: Английский

Ionic Conductivity Study of Antiperovskite Solid-State Electrolytes Based on Interpretable Machine Learning DOI
Shang Xiang, Shaowen Lu, Jiawei Li

et al.

ACS Applied Energy Materials, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 16, 2025

The development of high-performance all-solid-state ion batteries necessitates the design solid-state electrolytes (SSEs) with high ionic conductivity and excellent electrochemical stability. Antiperovskite (AP) X3BA, as electronically inverted derivative perovskite ABX3, has garnered significant attention in field energy storage due to its superior conductivity. However, relationship between their structure diffusion behavior warrants further investigation. In this work, we constructed a machine learning (ML) framework for predicting analyzing AP SSE, which encompasses data collection, feature selection, training various ML models. optimal model demonstrated an exceptional classification performance, achieving accuracy rate 94%. Furthermore, employed substitution method expand sample size from 168 150,000 orders magnitude. Based on expanded set, examined analyzed mechanisms underlying big perspective. findings reveal strong correlation atomic-scale characteristics at A-site. electronegativity, density, radius A-site are identified three most critical features influencing interpretable study enables high-precision prediction materials, provides insightful principles, significantly accelerates application SSEs.

Language: Английский

Citations

0

Determining whether biochar can effectively increase crop yields: A machine learning model development with imbalanced data DOI Creative Commons
Weidong Jiao, Kechao Li, Min Zhou

et al.

Environmental Technology & Innovation, Journal Year: 2025, Volume and Issue: unknown, P. 104154 - 104154

Published: March 1, 2025

Language: Английский

Citations

0

Machine Learning-Accelerated Exploration on Element Doping-Triggering Material Performance Improvement for Energy Conversion and Storage Applications DOI
Hao Wang, Y. S. Zhu, Jinliang Li

et al.

Journal of Materials Chemistry A, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

The prediction performances of machine learning in the field element-doped materials for energy conversion and storage applications are summarized.

Language: Английский

Citations

0

Matini-Net: Versatile Material Informatics Research Framework for Feature Engineering and Deep Neural Network Design DOI
Myeonghun Lee, Taehyun Park, Kyoungmin Min

et al.

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 21, 2024

In this study, we introduced Matini-Net, which is a versatile framework for feature engineering and automated architecture design materials informatics research using deep neural networks. Matini-Net provides the flexibility to feature-based, graph-based, combinations of these models, accommodating both single- multimodal model architectures. For validation, performed performance evaluation on MatBench benchmarking dataset five properties, targeting types regression architectures that can be designed Matini-Net. When applied each material property datasets, best various exhibited

Language: Английский

Citations

0