Article
Physics, Multidisciplinary
George Alexandru Nemnes, Nicolae Filipoiu, Valentin Sipica
Summary: The proposed workflow includes feature selection as a key step for optimizing research methods. Energy gaps of hybrid graphene-boron nitride nanoflakes were predicted using artificial neural networks, with training data obtained by associating structural information to the target quantity. Proper feature vector selection is crucial for accurate and efficient models.
Article
Crystallography
Sok- Tam, Pak-Kin Leong, Chi-Pui Tang, Weng-Hang Leong, Toshimori Sekine, Chi-Long Tang, Kuan-Vai Tam, U. Kin-Tak
Summary: By substituting the A site in P2(1)/c-A(CN)(2) and varying the lattice parameters, Al(CN)(2) and Si(CN)(2) were selected as candidate compounds with the lowest cohesive energy. The structures of the two candidates were optimized, and their phonon, electronic, and elastic properties were calculated. Si(CN)(2) showed potential for high-temperature and high-power applications due to its wide bandgap.
Article
Chemistry, Physical
Jakub Kocak, Eli Kraisler, Axel Schild
Summary: When a molecule dissociates, the Kohn-Sham (KS) and Pauli potentials form step structures, critical for describing dissociation and charge-transfer processes. The exact electron factorization (EEF) provides an explanation for these steps, showing they are a result of spatial electron entanglement and charge transfer. Additionally, two methods to reproduce the potentials during dissociation are proposed, offering insights into the encoding of many-electron effects in a one-electron theory.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2021)
Article
Engineering, Multidisciplinary
Kevin Linka, Amelie Schafer, Xuhui Meng, Zongren Zou, George Em Karniadakis, Ellen Kuhl
Summary: Understanding real-world dynamical phenomena is challenging, and machine learning has become the go-to technology for analyzing and making decisions based on these phenomena. However, traditional neural networks often ignore the fundamental laws of physics and fail to make accurate predictions. In this study, the combination of neural networks, physics informed modeling, and Bayesian inference is used to integrate data, physics, and uncertainties, improving the predictive potential of neural network models.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Martin Magris, Alexandros Iosifidis
Summary: The last decade has seen a growing interest in Bayesian learning, but its technicality and complexity in practical implementations have limited its widespread adoption. This survey introduces the principles and algorithms of Bayesian Learning for Neural Networks from a practical perspective, discussing standard and recent approaches for Bayesian inference. It also explores the use of manifold optimization as a state-of-the-art approach and provides pseudo-codes for implementation.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Statistics & Probability
Akifumi Okuno, Keisuke Yano
Summary: This article discusses the estimation of the generalization gap for overparameterized models, including neural networks. It shows that the functional variance characterizes the generalization gap even in overparameterized settings. To overcome the computational cost, the Langevin approximation of the functional variance (Langevin FV) is proposed, which only relies on the first-order gradient of the squared loss function. The method is demonstrated numerically on overparameterized linear regression and nonlinear neural network models.
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS
(2023)
Article
Automation & Control Systems
Ba-Hien Tran, Simone Rossi, Dimitrios Milios, Maurizio Filippone
Summary: This paper proposes a method of reasoning in neural networks using functional priors and matching the prior distributions with Gaussian processes. Experimental evidence shows that this approach improves performance significantly compared to other priors and approximate Bayesian deep learning methods.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Chemistry, Physical
Donghyeon Kang, Hyeon Yeong Lee, Joon-Ha Hwang, Sera Jeon, Dabin Kim, SeongMin Kim, Sang-Woo Kim
Summary: This study investigated the impact of molecular structure deformation on the triboelectric properties of PTFE induced by contact force using density functional theory (DFT). The results showed that deformation enhanced the negative triboelectric property of PTFE mainly due to the electron-deficient state of carbon atoms leading to enhanced local dipole.
Article
Energy & Fuels
Alessandro Brusaferri, Matteo Matteucci, Stefano Spinelli, Andrea Vitali
Summary: This work presents a novel approach to improve the trustworthiness of neural network based load forecasting systems by integrating predictive distributions and uncertainty sources. Experimental results demonstrate significant performance improvements in load forecasting.
Article
Physics, Multidisciplinary
Ludwig Winkler, Cesar Ojeda, Manfred Opper
Summary: This paper proposes a method to leverage the Bayesian uncertainty information encoded in parameter distributions to inform the learning procedure for Bayesian models. By deriving a Bayesian stochastic differential equation and applying stochastic optimal control, individually controlled learning rates are obtained for variational parameters. The resulting optimizer shows robustness to large learning rates and can adaptively and individually control the learning rates.
Article
Multidisciplinary Sciences
Boris Hanin, Alexander Zlokapa
Summary: This article investigates how the depth, width, and dataset size of neural networks jointly affect model quality and presents a complete solution in the case of linear networks. The study reveals the joint role of depth, width, and dataset size through asymptotic expansions of Meijer-G functions. It shows that linear networks make provably optimal predictions at infinite depth.
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
(2023)
Article
Chemistry, Physical
Cheng Cai, Weiqiang Tang, Chongzhi Qiao, Bo Bao, Peng Xie, Shuangliang Zhao, Honglai Liu
Summary: This article investigates the solvent effects on nucleophilic addition reactions in aqueous solution using the proposed multiscale reaction density functional theory (RxDFT) method. The results provide a better understanding of the mechanism and the solvent effect on chemical reactions, especially those occurring at solid-liquid interfaces.
GREEN ENERGY & ENVIRONMENT
(2022)
Article
Multidisciplinary Sciences
Xiaoxun Gong, He Li, Nianlong Zou, Runzhang Xu, Wenhui Duan, Yong Xu
Summary: The authors propose an E(3)-equivariant deep-learning framework for representing the density functional theory (DFT) Hamiltonian in material structures, which preserves Euclidean symmetry and allows for efficient electronic structure calculations.
NATURE COMMUNICATIONS
(2023)
Article
Physics, Applied
Andrew C. Antony, Dean Thelen, Nikolay Zhelev, Kaveh Adib, Robert G. Manley
Summary: The study reveals that contact charging between metals and hydroxylated SiO2 induces electronic states at the SiO2 surface, with the magnitude of these states varying with the type of metal and the density of surface defect states.
JOURNAL OF APPLIED PHYSICS
(2021)
Article
Chemistry, Physical
Aysenur Gencer, Sezgin Aydin, Ozge Surucu, Xiaotian Wang, Engin Deligoz, Gokhan Surucu
Summary: In this study, the hydrogen storage properties of Li-decorated Hf2CF2 MXene layer were investigated using first-principles calculations. The results show that the Li-decorated layer exhibits stable and convenient adsorption characteristics, making it a promising candidate for hydrogen storage applications.
APPLIED SURFACE SCIENCE
(2021)