Identifying flow defects in amorphous alloys using machine learning outlier detection methods

Tian, Liang; Fan, Yanqin; Li, Lin; Mousseau, Normand

doi:10.1016/j.scriptamat.2020.05.038

Cited by 34 publications

(11 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[142][143][144] Except for the energy related materials, so far, machine learning models and predictions have been utilized for accelerating the invention and research processes of amorphous and disordered materials of metallic glasses, 145,146 such as Co-V-Zr 147 and Cu-Zr. 148 As a whole, there are three main modes of applying the machine learning method to glassy material research, including data-driven machine learning, machine learning interatomic potential, and machine learning-driven AIMD simulation.…”

Section: Machine Learning For Materials Researchmentioning

confidence: 99%

Progress, challenges and perspectives of computational studies on glassy superionic conductors for solid-state batteries

Xia

2022

J. Mater. Chem. A

View full text Add to dashboard Cite

show abstract

Section: Machine Learning For Materials Researchmentioning

confidence: 99%

Progress, challenges and perspectives of computational studies on glassy superionic conductors for solid-state batteries

Xia

2022

J. Mater. Chem. A

View full text Add to dashboard Cite

show abstract

“…It is of note that a third route has appeared recently. The emerging machine-learning strategies represent a great advancement in this direction (Cubuk et al, 2015;Schoenholz et al, 2016;Schoenholz et al, 2017;Wang and Jain, 2019;Tian et al, 2020;Zhang et al, 2021), which have yielded an unprecedented accuracy in predicting local structural features and other dynamics in glasses (Fan et al, 2020;Wang et al, 2020;Yang et al, 2021). Despite its advantage in dealing with big data, the machine-learning model usually works as a black box, causing some puzzles in interpreting the datadriven results from a physically relevant perspective.…”

Section: Introductionmentioning

confidence: 99%

Ergodic Structural Diversity Predicts Dynamics in Amorphous Materials

Yang

Wang

2022

Front. Mater.

View full text Add to dashboard Cite

Identification of flexible local environments from a disordered medium has been a long-standing challenge. Here, we introduce a time-relevant structural Shannon entropy as a unique feature of the atomic-scale environment in glass, which is based on a metric of the time-invariant, or ergodic, and Voronoi structural diversity that an atom experiences during a sufficiently long-time thermal fluctuation. This new concept of time-relevant Shannon entropy simultaneously integrates the static topology and the vibrational feature such that it potentially probes all the possible configurational space in a sub-basin of the local potential energy landscape. This structural representation is not only capable of predicting the energy barrier of an elementary structural excitation but also demonstrates a robust correlation with the boson peak in metallic glasses, although the physical entity is defined from a purely structural aspect. The proposition, therefore, represents a successful demonstration of the physics-informed structure–property relationship in amorphous materials.

show abstract

“…Recently, the advancement of machine learning (ML) algorithms for handling various complex problems with nonlinearity or high dimensionality (Harrington et al, 2019;Tian et al, 2020) has led to STZ prediction studies based on ML algorithms, starting with Cubuk's pioneering study which defined "softness" using a support vector machine with the geometric features of atoms (Cubuk et al, 2015). In this line of studies, the predictive performance of a trained ML model was evaluated by metrics such as the AUC (area under the curve) of ROC (receiver operating characteristic) curve or with a probability plot associated with recall (Schoenholz et al, 2016;Cubuk et al, 2017;Wang and Jain, 2019;Wang et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Evaluating the predictive power of machine learning model for shear transformation in metallic glasses using metrics for an imbalanced dataset

Lee

Ryu

2022

Front. Mater.

View full text Add to dashboard Cite

Plastic deformation of metallic glasses, which show no long-range structural order, proceeds by shear transformation of a local group of atoms referred to as the shear transformation zone (STZ). Unlike crystalline solids, it is difficult to identify STZs and predict the onset of plasticity from a random atomic configuration under a given loading. Recently, significant efforts have been made to predict the shear transformation with initial atomic properties using machine learning. However, despite the class imbalance, where the atoms participating in shear transformation is much rarer compared to the others, few studies have explored the issue of the proper predictive metric choice, with most studies considering widely used metrics such as Recall or AUC in the machine learning community. Therefore, here we train a graph neural network that predicts the initially activated STZ and evaluate its predictive power using various metrics considered to be proper for handling imbalanced datasets. We find that the AUC value is significantly overestimated due to the class imbalance and too many atoms are misclassified as initial STZ, so other metrics such as the precision, f1, MCC, and AP indicate very low predictive power close to zero. Additionally, we reveal that the predictive performance changes significantly over the threshold value of non-affine displacement, above which an atom is classified as the initially activated STZ, due to the change in the degree of class imbalance. Our study implies that it is crucial to use an identical threshold for this type of classification (i.e., the class ratio) for a fair assessment of ML models adapted in different studies and to holistically evaluate the predictive performance based on various metrics.

show abstract

Identifying flow defects in amorphous alloys using machine learning outlier detection methods

Cited by 34 publications

References 53 publications

Progress, challenges and perspectives of computational studies on glassy superionic conductors for solid-state batteries

Progress, challenges and perspectives of computational studies on glassy superionic conductors for solid-state batteries

Ergodic Structural Diversity Predicts Dynamics in Amorphous Materials

Evaluating the predictive power of machine learning model for shear transformation in metallic glasses using metrics for an imbalanced dataset

Contact Info

Product

Resources

About