Improving Symbolic Regression for Predicting Materials Properties with Iterative Variable Selection

Guo, Zhenghong; Hu, Shunbo; Han, Zhongkang; Ouyang, Runhai

doi:10.1021/acs.jctc.2c00281

Cited by 19 publications

(10 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Symbolic regression is a genetic programming-based machine learning technique designed to identify an underlying mathematical expression 87,88 . It first builds a stochastic formula to represent the relationship between known independent and dependent variables to predict data.…”

Section: Modeling Algorithms For Small Datamentioning

confidence: 99%

Small data machine learning in materials science

et al. 2023

npj Comput Mater

127

View full text Add to dashboard Cite

This review discussed the dilemma of small data faced by materials machine learning. First, we analyzed the limitations brought by small data. Then, the workflow of materials machine learning has been introduced. Next, the methods of dealing with small data were introduced, including data extraction from publications, materials database construction, high-throughput computations and experiments from the data source level; modeling algorithms for small data and imbalanced learning from the algorithm level; active learning and transfer learning from the machine learning strategy level. Finally, the future directions for small data machine learning in materials science were proposed.

show abstract

Section: Modeling Algorithms For Small Datamentioning

confidence: 99%

Small data machine learning in materials science

et al. 2023

npj Comput Mater

127

View full text Add to dashboard Cite

show abstract

“…More broadly, symbolic regression corresponds to the ansatz that a relatively simple combination of mathematical functions describes the behavior. There are a variety of applications of symbolic regression methods to problems in chemistry and materials science. , In practice, symbolic regression is often combined with various feature selection methods, with examples including VS-SISSO and transformer-based approaches for symbolic regression …”

Section: Recommendations Toward ML For Exceptional Materialsmentioning

confidence: 99%

“…There are a variety of applications of symbolic regression methods to problems in chemistry 150 and materials science. 151,152 In practice, symbolic regression is often combined with various feature selection methods, with examples including VS-SISSO 153 and transformer-based approaches for symbolic regression. 154 Emphasizing the qualitative direction has consequences for the design of HTE systems.…”

Section: Model Will Hinder Discovering Cuprate Superconductors)mentioning

confidence: 99%

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science

Schrier,

Norquist,

Buonassisi

et al. 2023

J. Am. Chem. Soc.

View full text Add to dashboard Cite

Exceptional molecules and materials with one or more extraordinary properties are both technologically valuable and fundamentally interesting, because they often involve new physical phenomena or new compositions that defy expectations. Historically, exceptionality has been achieved through serendipity, but recently, machine learning (ML) and automated experimentation have been widely proposed to accelerate target identification and synthesis planning. In this Perspective, we argue that the data-driven methods commonly used today are well-suited for optimization but not for the realization of new exceptional materials or molecules. Finding such outliers should be possible using ML, but only by shifting away from using traditional ML approaches that tweak the composition, crystal structure, or reaction pathway. We highlight case studies of high-T c oxide superconductors and superhard materials to demonstrate the challenges of ML-guided discovery and discuss the limitations of automation for this task. We then provide six recommendations for the development of ML methods capable of exceptional materials discovery: (i) Avoid the tyranny of the middle and focus on extrema; (ii) When data are limited, qualitative predictions that provide direction are more valuable than interpolative accuracy; (iii) Sample what can be made and how to make it and defer optimization; (iv) Create room (and look) for the unexpected while pursuing your goal; (v) Try to fill-in-the-blanks of input and output space; (vi) Do not confuse human understanding with model interpretability. We conclude with a description of how these recommendations can be integrated into automated discovery workflows, which should enable the discovery of exceptional molecules and materials.

show abstract

“…In order to improve the interpretability and provide scientific information, alternative machine learning algorithms are such as symbolic regression have been applied to design relevant hybrid descriptors in chemically intuitive mathematical forms. Symbolic regression is a regression analysis that looks across the space of mathematical expressions to identify the model that most accurately and simply fits a given dataset [52][53][54]. It avoids imposing previous assumptions and infers the model from the data, whereas conventional regression approaches aim to optimize the parameters for a pre-specified model structure.…”

Section: Introductionmentioning

confidence: 99%

“…Ouyang et al, proposed a SISSO-based symbolic regression approach for discovering descriptors for material properties, which is based on the compressed-sensing dimensionality reduction [56]. The symbolic regression is integrated with iterative variable selection to efficiently handle hundreds of input features [52]. This approach has been widely applied for descriptor design in many applications (e.g.…”

Section: Introductionmentioning

confidence: 99%

First-principles and machine learning investigation on A4BX6 halide perovskites

Zheng,

Huang,

Zhang

2023

Modelling Simul. Mater. Sci. Eng.

View full text Add to dashboard Cite

The A4BX6 molecular halide perovskites have received attention owing to their interesting optoelectronic properties at the molecular scale; however, a comprehensive dataset of their atomic structures and electronic properties and associated data-driven investigation are still unavailable now, which makes it difficult for inverse materials design for semiconductor applications (e.g. wide band gap semiconductor). In this manuscript, we employ data-driven methods to predict band gaps of A4BX6 molecular halide perovskites via machine learning. A large virtual design database including 246 904 A4BX6 perovskite samples is predicted via machine learning, based on the model trained using 2740 first-principles results of A4BX6 molecular halide perovskites. In addition, symbolic regression-based machine learning is employed to identify more physically intuitive descriptors based on the starting first-principles dataset of A4BX6 molecular halide perovskites. In addition, different ranking methods are employed to offer a comprehensive feature importance analysis for the halide perovskite materials. This study highlights the efficacy of machine learning-assisted compositional design of A4BX6 perovskites, and the multi-dimensional database established here is valuable for future experimental validation toward perovskite-based wide band gap semiconductor materials.

show abstract

Improving Symbolic Regression for Predicting Materials Properties with Iterative Variable Selection

Cited by 19 publications

References 41 publications

Small data machine learning in materials science

Small data machine learning in materials science

In Pursuit of the Exceptional: Research Directions for Machine Learning in Chemical and Materials Science

First-principles and machine learning investigation on A4BX6 halide perovskites

Contact Info

Product

Resources

About