Mapping mesoscopic phase evolution during E-beam induced transformations via deep learning of atomically resolved images

Vasudevan, Rama K.; Laanait, Nouamane; Ferragut, Erik M.; Wang, Kai; Geohegan, David B.; Xiao, Kai; Ziatdinov, Maxim; Jesse, Stephen; Dyck, Ondrej; Kalinin, Sergei V.

doi:10.1038/s41524-018-0086-7

Cited by 46 publications

(53 citation statements)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The symmetry of the probed nearsurface region is to be described by an orthogonal projection of a subperiodic (3D) layer group. 14 This author's research group reproduced this particular result also on the basis of an information theory based plane symmetry group classification [16] since that STM image was freely downloadable from the on-line support material of [46]. The symmetrized version of that experimental STM image in plane symmetry group h31m revealed both carbon atoms in the primitive p3m1 unit cell [16,73].…”

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 78%

“…Due to the nature 13 of the contrast in STM images, their plane symmetry is not necessarily the one which is obtained by an orthogonal projection from the 3D space group symmetry, as listed in [1] for several high symmetry directions. Graphite [50] served as sample in both of the classifications of experimental STM images in [46].…”

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 99%

“…There were incorrect usages of crystallographic standard terms for the 2D Bravais lattice types [1,3,44] as well as misleading statements about the nature of some of their samples [50], the existence of clearly discernable Moiré fringes in the amplitude map of the discrete Fourier transform of one of their images [16], and a figure which implied that classification probabilities by their DCNN could be negative or in excess of 100 % in the original version of the paper by Vasudevan and co-workers [46]. The crystallographic misnomers have been corrected to a large extent in the current (December 18, 2018) on-line version of their paper, i.e.…”

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 99%

“…Deep convolutional neural networks were recently also applied to synthetic "images" that represent crystallographic symmetry information in 3D [47,48]. Just as in the case of the neural network for 2D Bravais lattice type classifications [46] (as discussed in the fifth section of this paper), pseudosymmetries present challenges to these networks in conjunction with generalized crystal structure data recording and processing noise [44], but were neglected in both of these studies. Because noisy images with G-AIC based labels were not part of the training sets of these neural networks that worked with crystallographic 3D data [47,48] (as judged from the available information in these papers), the classification performance of these networks must also be limited.…”

Section: Appendix: Classifications Of 3d Crystals By Neural Network mentioning

confidence: 99%

“…An appendix discusses the state of the art in classifications of selected aspects of crystal symmetries in 3D by means of DCNNs [47,48] and provides the basis for comparisons of their performance to that of the 2D translation symmetry classifier of [46] (as discussed in the fifth section of this paper). A classification of crystalline compounds by such machines that is aided by the morphologies of individual crystals [49] is also mentioned in the appendix.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

On Classification Approaches for Crystallographic Symmetries of Noisy 2D Periodic Patterns

Moeck

2019

IEEE Trans. Nanotechnology

View full text Add to dashboard Cite

The existing types of classification approaches for the crystallographic symmetries of patterns that are more or less periodic in two dimensions (2D) are reviewed. Their relative performance is evaluated in a qualitative manner. Pseudosymmetries of different kinds are discussed as they present severe challenges to most classification approaches when noise levels are moderate to high. The author's information theory based approaches utilize digital images and geometric Akaike Information Criteria. They perform well in the presence of pseudo-symmetries and turn out to be the only ones that allow for fully objective (completely researcher independent) and generalized noise level dependent classifications of the full range of crystallographic symmetries, i.e. Bravais lattice type, Laue class, and plane symmetry group, of noisy real-world images. His method's identification of the plane symmetry group that can with the highest likelihood be assigned to a noisy 2D periodic image enables the most meaningful crystallographic averaging in the spatial frequency domain. This kind of averaging suppresses generalized noise much more effectively than traditional Fourier filtering. Taking account of the fact that it is fundamentally unsound to assign an abstract mathematical concept such as a single symmetry type, class, or group with 100 % certainty to a more or less 2D periodic record of a noisy real-world imaging experiment that involved a real-world sample, the author's information theory based approaches to crystallographic symmetry classifications deliver probabilistic (rather than definitive) classifications. Recent applications of deep convolutional neural networks (DCNNs) to classifications of crystallographic translation symmetry types in 2D and crystals in three dimensions (3D) are discussed as these "correlation detection and optimization" machines deliver probabilistic classifications by other -non-analytical -means. The discussed DCNN classifications ignore the fact that many crystallographic symmetries are hierarchic, i.e. that the classification classes are often non-disjoint. They are currently also incapable of dealing with pseudo-symmetries. DCNNs for classifications of crystals in 3D are discussed separately in an appendix.

show abstract

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 78%

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 99%

Section: D Translation Symmetry Classification By a Neural Networmentioning

confidence: 99%

Section: Appendix: Classifications Of 3d Crystals By Neural Network mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

On Classification Approaches for Crystallographic Symmetries of Noisy 2D Periodic Patterns

Moeck

2019

IEEE Trans. Nanotechnology

View full text Add to dashboard Cite

show abstract

Toward Excellence of Electrocatalyst Design by Emerging Descriptor‐Oriented Machine Learning

Liu

Luo

Wang

et al. 2022

Adv Funct Materials

View full text Add to dashboard Cite

(1 of 25)development, it is still far from meeting the increased demand. [3] The emergence of artificial intelligence (AI) has fundamentally changed the situation, which has significantly accelerated the discovery process, owing to greatly improved algorithms and developments in data science. [4] Machine learning (ML), a simple and practical AI framework based on computer and statistical science, is used to develop algorithms to learn from historic data without being explicitly programmed to obtain specific results. [5] It can investigate relationships that are hard to clearly and definitely model mathematically, providing insights for new scientific advancements related to highly complex with many uncertain twisted together factors. [6] There are usually three factors that govern the learning and prediction process of an ML: algorithms, data/database, and descriptors. [4] The algorithms involve data extraction, data filing, and propagation from mathematical derivation. [5,7] The data can be derived not only from experiments but also from theoretical calculations. [4] A number of databases based on experiments [8] and calculations [9] have already been established. The descriptors depend to a large extent on the predicted material or properties. Based on the algorithm, databases, and descriptors, the ML applications have been successfully implemented to support various energy materials with analysis tools (e.g., Python-based SciKit-Learn [10] and TensorFlow [11] ) in combination with workflow management tools (e.g., ASE [12] and Atomate [13] ). However, the prediction accuracy depends highly on the descriptors, as descriptors have a certain uniqueness for various materials and properties as long as the algorithm is selected correctly and the data set is complete. [14] For catalysis, the descriptors contain the essence from the physicochemical nature. Based on effective descriptors, ML can uncover the relationship bridging structure and its activity, selectivity, and stability. [5,15] Thus, suitable descriptors must be established to understand the structure-activity relationship. Although many efforts have been made to accelerate the rational design of homogeneous catalysts, [7b,16] heterogeneous catalysts, [7b,16b,17] and electrocatalysis, [3,18] the development of ML-assisted real catalysts is still in its infancy. Despite these considerable research efforts, the lack of universal selection tactics for descriptors bridging the gap between activity and structures impedes the application

show abstract

SIGMA: Spectral Interpretation using Gaussian Mixtures and Autoencoder

Tung

Sheikh

Ball

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Mapping mesoscopic phase evolution during E-beam induced transformations via deep learning of atomically resolved images

Cited by 46 publications

References 58 publications

On Classification Approaches for Crystallographic Symmetries of Noisy 2D Periodic Patterns

On Classification Approaches for Crystallographic Symmetries of Noisy 2D Periodic Patterns

Toward Excellence of Electrocatalyst Design by Emerging Descriptor‐Oriented Machine Learning

SIGMA: Spectral Interpretation using Gaussian Mixtures and Autoencoder

Contact Info

Product

Resources

About