Nouvelles Slovènes

Mihailovich, Vasa D.

doi:10.2307/40125193

Search citation statements

Order By: Relevance

Paper Sections

Select...

Introduction1

Citation Types

Supporting

Mentioning

Contrasting

Year Published

2020

Publication Types

Select...

Preprint1

Relationship

Self Cite0

Independent1

Authors

Journals

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, this is not true in all cases; an extreme example of this is element doping. Dopants can be present in very small amounts in a compound, but can have a significant impact on its electrical, 23,26,27 mechanical, 20,[28][29][30] and thermal properties. [31][32][33][34] In the case of a typical CBFV approach which uses the weighted average of element properties as a feature, the signal from dopant elements would not significantly change the vector representation of a compound.…”

Section: Introductionmentioning

confidence: 99%

Compositionally-Restricted Attention-Based Network for Materials Property Prediction

Wang¹,

Kauwe²,

Murdock³

et al. 2020

Preprint

View full text Add to dashboard Cite

<div>In this paper, we demonstrate a novel application of the Transformer self-attention mechanism. Our network, the Compositionally-Restricted Attention-Based network, referred to as CrabNet, explores the area of structure-agnostic materials property predictions when only a chemical formula is provided.</div><div>Our results show that CrabNet's performance matches or exceeds current best practice methods on nearly all of 28 total benchmark datasets. We also demonstrate how CrabNet's architecture lends itself towards model interpretability by showing different visualization approaches that are made possible by CrabNet's design.</div><div>We feel confident that CrabNet, and its attention-based framework, will be of keen interest to future materials informatics researchers. </div>

show abstract