Physics-inspired machine learning of localized intensive properties

Chen, Ke; Künkel, Christian; Cheng, Bingqing; Margraf, Johannes T.

doi:10.1039/d3sc00841j

Cited by 8 publications

(14 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This task is typically handled with simple fixed pooling functions like sum, average, or maximum. Despite their appealing simplicity, there are growing concerns regarding the representational power of this class of functions 26 , 27 . In the following section, we also discuss the concurrently developed orbital-weighted average (OWA), a physics-based method designed specifically for orbital properties and which also seeks to improve upon the standard pooling operators by exploiting the local and intensive character of the target property 27 .…”

Section: Introductionmentioning

confidence: 99%

“…Despite their appealing simplicity, there are growing concerns regarding the representational power of this class of functions 26 , 27 . In the following section, we also discuss the concurrently developed orbital-weighted average (OWA), a physics-based method designed specifically for orbital properties and which also seeks to improve upon the standard pooling operators by exploiting the local and intensive character of the target property 27 . Buterez et al also highlighted the lacklustre performance of standard pooling functions in a variety of settings, particularly on challenging molecular properties 28 .…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modelling local and general quantum mechanical properties with attention-based pooling

Buterez,

Janet,

Kiddle

et al. 2023

Commun Chem

View full text Add to dashboard Cite

Atom-centred neural networks represent the state-of-the-art for approximating the quantum chemical properties of molecules, such as internal energies. While the design of machine learning architectures that respect chemical principles has continued to advance, the final atom pooling operation that is necessary to convert from atomic to molecular representations in most models remains relatively undeveloped. The most common choices, sum and average pooling, compute molecular representations that are naturally a good fit for many physical properties, while satisfying properties such as permutation invariance which are desirable from a geometric deep learning perspective. However, there are growing concerns that such simplistic functions might have limited representational power, while also being suboptimal for physical properties that are highly localised or intensive. Based on recent advances in graph representation learning, we investigate the use of a learnable pooling function that leverages an attention mechanism to model interactions between atom representations. The proposed pooling operation is a drop-in replacement requiring no changes to any of the other architectural components. Using SchNet and DimeNet++ as starting models, we demonstrate consistent uplifts in performance compared to sum and mean pooling and a recent physics-aware pooling operation designed specifically for orbital energies, on several datasets, properties, and levels of theory, with up to 85% improvements depending on the specific task.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Modelling local and general quantum mechanical properties with attention-based pooling

Buterez,

Janet,

Kiddle

et al. 2023

Commun Chem

View full text Add to dashboard Cite

show abstract

“…In addition, one can sometimes obtain physically unreasonable results. [ 103 , 104 , 105 ] As a result, neural networks and deep learning methods, which can be highly accurate, often yield results that are poorly interpretable. Fortunately, recent developments in so‐called “physics‐based” deep learning offer opportunities to construct robust and physically meaningful models from which interpretable information can be extracted.…”

Section: A Way Forwardmentioning

confidence: 99%

“…Fortunately, recent developments in so‐called “physics‐based” deep learning offer opportunities to construct robust and physically meaningful models from which interpretable information can be extracted. [ 98 , 99 , 100 , 101 , 102 , 103 ]…”

Section: A Way Forwardmentioning

confidence: 99%

Leave No Photon Behind: Artificial Intelligence in Multiscale Physics of Photocatalyst and Photoreactor Design

Loh,

Wang,

Mohan

et al. 2024

Advanced Science

View full text Add to dashboard Cite

Although solar fuels photocatalysis offers the promise of converting carbon dioxide directly with sunlight as commercially scalable solutions have remained elusive over the past few decades, despite significant advancements in photocatalysis band‐gap engineering and atomic site activity. The primary challenge lies not in the discovery of new catalyst materials, which are abundant, but in overcoming the bottlenecks related to material‐photoreactor synergy. These factors include achieving photogeneration and charge‐carrier recombination at reactive sites, utilizing high mass transfer efficiency supports, maximizing solar collection, and achieving uniform light distribution within a reactor. Addressing this multi‐dimensional problem necessitates harnessing machine learning techniques to analyze real‐world data from photoreactors and material properties. In this perspective, the challenges are outlined associated with each bottleneck factor, review relevant data analysis studies, and assess the requirements for developing a comprehensive solution that can unlock the full potential of solar fuels photocatalysis technology. Physics‐informed machine learning (or Physics Neural Networks) may be the key to advancing this important area from disparate data towards optimal reactor solutions.

show abstract

“…Eine rigorose Behandlung solcher Eigenschaften erfordert deshalb spezielle ML-Algorithmen, welche das physikalisch korrekte Skalierungsverhalten gewährleisten. 9) Interessanterweise führen solche methodischen Entwicklungen speziell dann zu Verbesserungen, wenn wenig Daten zur Verfügung stehen. Dies erlaubt es zu einem gewissen Maß, auch jenseits des Trainingssatzes zu extrapolieren.…”

Section: Vorhersage Elektronischer Eigenschaftenunclassified

Trendbericht Theoretische Chemie 2023 (1/3): Maschinelles Lernen jenseits von Potenzialenergiehyperflächen

Margraf

2023

Nachr Chem

View full text Add to dashboard Cite

Maschinelles Lernen: Bisher zielten Anwendungen vor allem auf das Lernen von Potenzialenergiehyperflächen, derzeit zeigen sich viele neue Forschungsrichtungen. Molekülschwingungen in der Thermodynamik: Berechnung in Theorie und Simulation. Ultrakurze Laserpulse zum Beobachten molekularer Dynamik: Da die Bewegung von Elektronen genauso schnell ist wie die zeitliche Ausdehnung der Pulse, sind die Pulseigenschaften bei Simulationen explizit zu berücksichtigen.

show abstract

Physics-inspired machine learning of localized intensive properties

Abstract: Machine learning (ML) has been widely applied to chemical property prediction, most prominently for the energies and forces in molecules and materials. The strong interest in predicting energies in particular...

Cited by 8 publications

References 56 publications

Modelling local and general quantum mechanical properties with attention-based pooling

Modelling local and general quantum mechanical properties with attention-based pooling

Leave No Photon Behind: Artificial Intelligence in Multiscale Physics of Photocatalyst and Photoreactor Design

Trendbericht Theoretische Chemie 2023 (1/3): Maschinelles Lernen jenseits von Potenzialenergiehyperflächen

Contact Info

Product

Resources

About