English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

We introduce a multi-tasking graph convolutional neural network, HydraGNN, tosimultaneously predict both global and atomic physical properties anddemonstrate with ferromagnetic materials. We train HydraGNN on an open-sourceab initio density functional theory (DFT) dataset for iron-platinum (FePt) witha fixed body centered tetragonal (BCT) lattice structure and fixed volume tosimultaneously predict the mixing enthalpy (a global feature of the system),the atomic charge transfer, and the atomic magnetic moment acrossconfigurations that span the entire compositional range. By taking advantage ofunderlying physical correlations between material properties, multi-tasklearning (MTL) with HydraGNN provides effective training even with modestamounts of data. Moreover, this is achieved with just one architecture insteadof three, as required by single-task learning (STL). The first convolutionallayers of the HydraGNN architecture are shared by all learning tasks andextract features common to all material properties. The following layersdiscriminate the features of the different properties, the results of which arefed to the separate heads of the final layer to produce predictions. Numericalresults show that HydraGNN effectively captures the relation between theconfigurational entropy and the material properties over the entirecompositional range. Overall, the accuracy of simultaneous MTL predictions iscomparable to the accuracy of the STL predictions. In addition, thecomputational cost of training HydraGNN for MTL is much lower than the originalDFT calculations and also lower than training separate STL models for eachproperty.

Multi-task graph neural networks for simultaneous prediction of global and atomic properties in ferromagnetic systems