skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.


Title: Enhancing material property prediction with ensemble deep graph convolutional networks
Machine learning (ML) models have emerged as powerful tools for accelerating materials discovery and design by enabling accurate predictions of properties from compositional and structural data. These capabilities are vital for developing advanced technologies across fields such as energy, electronics, and biomedicine, potentially reducing the time and resources needed for new material exploration and promoting rapid innovation cycles. Recent efforts have focused on employing advanced ML algorithms, including deep learning-based graph neural networks, for property prediction. Additionally, ensemble models have proven to enhance the generalizability and robustness of ML and Deep Learning (DL). However, the use of such ensemble strategies in deep graph networks for material property prediction remains underexplored. Our research provides an in-depth evaluation of ensemble strategies in deep learning-based graph neural network, specifically targeting material property prediction tasks. By testing the Crystal Graph Convolutional Neural Network (CGCNN) and its multitask version, MT-CGCNN, we demonstrated that ensemble techniques, especially prediction averaging, substantially improve precision beyond traditional metrics for key properties like formation energy per atom ( Δ E f ) , band gap ( E g ) , density ( ρ ) , equivalent reaction energy per atom ( E rxn,atom ) , energy per atom ( E atom ) and atomic density ( ρ atom ) in 33,990 stable inorganic materials. These findings support the broader application of ensemble methods to enhance predictive accuracy in the field.  more » « less
Award ID(s):
1650474
PAR ID:
10581010
Author(s) / Creator(s):
; ; ; ;
Publisher / Repository:
Frontiers
Date Published:
Journal Name:
Frontiers in Materials
Volume:
11
ISSN:
2296-8016
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract In this paper, we establish the existence of a bounded, linear extension operator T : L 2 , p ( E ) L 2 , p ( R 2 ) $$T :{L}^{2,p}\left(E\right)\to {L}^{2,p}\left({\mathbb{R}}^{2}\right)$$when 1 <p< 2 andEis a finite subset of R 2 $${\mathbb{R}}^{2}$$contained in a line. 
    more » « less
  2. We present a measurement of the branching fraction and fraction of longitudinal polarization of B 0 ρ + ρ decays, which have two π 0 ’s in the final state. We also measure time-dependent C P violation parameters for decays into longitudinally polarized ρ + ρ pairs. This analysis is based on a data sample containing ( 387 ± 6 ) × 10 6 ϒ ( 4 S ) mesons collected with the Belle II detector at the SuperKEKB asymmetric-energy e + e collider in 2019–2022. We obtain B ( B 0 ρ + ρ ) = ( 2.8 9 0.22 + 0.23 0.27 + 0.29 ) × 10 5 , f L = 0.92 1 0.025 + 0.024 0.015 + 0.017 , S = 0.26 ± 0.19 ± 0.08 , and C = 0.02 ± 0.1 2 0.05 + 0.06 , where the first uncertainties are statistical and the second are systematic. We use these results to perform an isospin analysis to constrain the Cabibbo-Kobayashi-Maskawa angle ϕ 2 and obtain two solutions; the result consistent with other Standard Model constraints is ϕ 2 = ( 92.6 4.7 + 4.5 ) ° . Published by the American Physical Society2025 
    more » « less
  3. Abstract We show that the affine vertex superalgebra V k ( o s p 1 | 2 n ) V^{k}(\mathfrak{osp}_{1|2n})at generic level 𝑘 embeds in the equivariant 𝒲-algebra of s p 2 n \mathfrak{sp}_{2n}times 4 n 4nfree fermions.This has two corollaries:(1) it provides a new proof that, for generic 𝑘, the coset Com ( V k ( s p 2 n ) , V k ( o s p 1 | 2 n ) ) \operatorname{Com}(V^{k}(\mathfrak{sp}_{2n}),V^{k}(\mathfrak{osp}_{1|2n}))is isomorphic to W ( s p 2 n ) \mathcal{W}^{\ell}(\mathfrak{sp}_{2n})for = ( n + 1 ) + ( k + n + 1 ) / ( 2 k + 2 n + 1 ) \ell=-(n+1)+(k+n+1)/(2k+2n+1), and(2) we obtain the decomposition of ordinary V k ( o s p 1 | 2 n ) V^{k}(\mathfrak{osp}_{1|2n})-modules into V k ( s p 2 n ) W ( s p 2 n ) V^{k}(\mathfrak{sp}_{2n})\otimes\mathcal{W}^{\ell}(\mathfrak{sp}_{2n})-modules.Next, if 𝑘 is an admissible level and ℓ is a non-degenerate admissible level for s p 2 n \mathfrak{sp}_{2n}, we show that the simple algebra L k ( o s p 1 | 2 n ) L_{k}(\mathfrak{osp}_{1|2n})is an extension of the simple subalgebra L k ( s p 2 n ) W ( s p 2 n ) L_{k}(\mathfrak{sp}_{2n})\otimes{\mathcal{W}}_{\ell}(\mathfrak{sp}_{2n}).Using the theory of vertex superalgebra extensions, we prove that the category of ordinary L k ( o s p 1 | 2 n ) L_{k}(\mathfrak{osp}_{1|2n})-modules is a semisimple, rigid vertex tensor supercategory with only finitely many inequivalent simple objects.It is equivalent to a certain subcategory of W ( s p 2 n ) \mathcal{W}_{\ell}(\mathfrak{sp}_{2n})-modules.A similar result also holds for the category of Ramond twisted modules.Due to a recent theorem of Robert McRae, we get as a corollary that categories of ordinary L k ( s p 2 n ) L_{k}(\mathfrak{sp}_{2n})-modules are rigid. 
    more » « less
  4. Abstract We extend the Calderón–Zygmund theory for nonlocal equations tostrongly coupled system of linear nonlocal equations A s u = f {\mathcal{L}^{s}_{A}u=f}, where the operator A s {\mathcal{L}^{s}_{A}}is formally given by A s u = n A ( x , y ) | x - y | n + 2 s ( x - y ) ( x - y ) | x - y | 2 ( u ( x ) - u ( y ) ) 𝑑 y . \mathcal{L}^{s}_{A}u=\int_{\mathbb{R}^{n}}\frac{A(x,y)}{|x-y|^{n+2s}}\frac{(x-%y)\otimes(x-y)}{|x-y|^{2}}(u(x)-u(y))\,dy. For 0 < s < 1 {0<1}and A : n × n {A:\mathbb{R}^{n}\times\mathbb{R}^{n}\to\mathbb{R}}taken to be symmetric and serving asa variable coefficient for the operator, the system under consideration is the fractional version of the classical Navier–Lamé linearized elasticity system. The study of the coupled system of nonlocal equations is motivated by its appearance in nonlocal mechanics, primarily in peridynamics. Our regularity result states that if A ( , y ) {A(\,\cdot\,,y)}is uniformly Holder continuous and inf x n A ( x , x ) > 0 {\inf_{x\in\mathbb{R}^{n}}A(x,x)>0}, then for f L loc p {f\in L^{p}_{\rm loc}}, for p 2 {p\geq 2}, the solution vector u H loc 2 s - δ , p {u\in H^{2s-\delta,p}_{\rm loc}}for some δ ( 0 , s ) {\delta\in(0,s)}. 
    more » « less
  5. Abstract Ultra-high-energy cosmic rays (UHECRs), particles characterized by energies exceeding 1018eV, are generally believed to be accelerated electromagnetically in high-energy astrophysical sources. One promising mechanism of UHECR acceleration is magnetized turbulence. We demonstrate from first principles, using fully kinetic particle-in-cell simulations, that magnetically dominated turbulence accelerates particles on a short timescale, producing a power-law energy distribution with a rigidity-dependent, sharply defined cutoff well approximated by the form f cut E , E cut = sech ( E / E cut ) 2 . Particle escape from the turbulent accelerating region is energy dependent, withtesc∝E−δandδ∼ 1/3. The resulting particle flux from the accelerator follows dN / dEdt E s sech ( E / E cut ) 2 , withs∼ 2.1. We fit the Pierre Auger Observatory’s spectrum and composition measurements, taking into account particle interactions between acceleration and detection, and show that the turbulence-associated energy cutoff is well supported by the data, with the best-fitting spectral index being s = 2.1 0.13 + 0.06 . Our first-principles results indicate that particle acceleration by magnetically dominated turbulence may constitute the physical mechanism responsible for UHECR acceleration. 
    more » « less