检索结果-内蒙古大学图书馆

Position: Bayesian Deep Learning is Needed in the Age of Large-Scale AI 41

学校读者我要写书评

暂无评论

Position: Bayesian Deep Learning is Needed in the Age of Lar...

41st International Conference on Machine Learning, ICML 2024

作者： Papamarkou, Theodore Skoularidou, Maria Palla, Konstantina Aitchison, Laurence Arbel, Julyan Dunson, David Filippone, Maurizio Fortuin, Vincent Hennig, Philipp Hernández-Lobato, José Miguel Hubin, Aliaksandr Immer, Alexander Karaletsos, Theofanis Khan, Mohammad Emtiyaz Kristiadi, Agustinus Li, Yingzhen Mandt, Stephan Nemeth, Christopher Osborne, Michael A. Rudner, Tim G.J. Rügamer, David Teh, Yee Whye Welling, Max Wilson, Andrew Gordon Zhang, Ruqi Department of Mathematics The University of Manchester Manchester United Kingdom Eric and Wendy Schmidt Center Broad Institute of MIT and Harvard Cambridge United States Spotify London United Kingdom Computational Neuroscience Unit University of Bristol Bristol United Kingdom Centre Inria de l'Université Grenoble Alpes Grenoble France Department of Statistical Science Duke University United States Statistics Program KAUST Saudi Arabia Helmholtz AI Munich Germany Department of Computer Science Technical University of Munich Munich Germany Munich Center for Machine Learning Munich Germany Tübingen AI Center University of Tübingen Tübingen Germany Department of Engineering University of Cambridge Cambridge United Kingdom Department of Mathematics University of Oslo Oslo Norway Bioinformatics and Applied Statistics Norwegian University of Life Sciences Ås Norway Department of Computer Science ETH Zurich Switzerland Chan Zuckerberg Initiative CA United States Center for Advanced Intelligence Project RIKEN Tokyo Japan Vector Institute Toronto Canada Department of Computing Imperial College London London United Kingdom Department of Computer Science UC Irvine Irvine United States Department of Mathematics and Statistics Lancaster University Lancaster United Kingdom Department of Engineering Science University of Oxford Oxford United Kingdom Center for Data Science New York University New York United States Department of Statistics LMU Munich Munich Germany DeepMind London United Kingdom Department of Statistics University of Oxford Oxford United Kingdom Informatics Institute University of Amsterdam Amsterdam Netherlands Courant Institute of Mathematical Sciences Center for Data Science Computer Science Department New York University New York United States Department of Computer Science Purdue University West Lafayette United States

In the current landscape of deep learning research, there is a predominant emphasis on achieving high predictive accuracy in supervised tasks involving large image and language datasets. However, a broader perspective reveals a multitude of overlooked metrics, tasks, and data types, such as uncertainty, active and continual learning, and scientific data, that demand attention. Bayesian deep learning (BDL) constitutes a promising avenue, offering advantages across these diverse settings. This paper posits that BDL can elevate the capabilities of deep learning. It revisits the strengths of BDL, acknowledges existing challenges, and highlights some exciting research avenues aimed at addressing these obstacles. Looking ahead, the discussion focuses on possible ways to combine large-scale foundation models with BDL to unlock their full potential. Copyright 2024 by the author(s)

关键词： Contrastive Learning

Continuous Spatiotemporal Transformers

学校读者我要写书评

暂无评论

arXiv 2023年

作者： de Oliveira Fonseca, Antonio Henrique Zappala, Emanuele Caro, Josue Ortega van Dijk, David Interdepartmental Neuroscience Program Yale University New HavenCT United States Department of Computer Science Yale University New HavenCT United States Department of Neuroscience Yale University New HavenCT United States Wu Tsai Institute Yale University New HavenCT United States Yale University New HavenCT United States Interdepartmental Program in Computational Biology & Bioinformatics Yale University New HavenCT United States

Modeling spatiotemporal dynamical systems is a fundamental challenge in machine learning. Transformer models have been very successful in NLP and computer vision where they provide interpretable representations of data. However, a limitation of transformers in modeling continuous dynamical systems is that they are fundamentally discrete time and space models and thus have no guarantees regarding continuous sampling. To address this challenge, we present the Continuous Spatiotemporal Transformer (CST), a new transformer architecture that is designed for modeling of continuous systems. This new framework guarantees a continuous and smooth output via optimization in Sobolev space. We benchmark CST against traditional transformers as well as other spatiotemporal dynamics modeling methods and achieve superior performance in a number of tasks on synthetic and real systems, including learning brain dynamics from calcium imaging data. © 2023, CC BY-NC-SA.

关键词： Sobolev spaces

Reconstructing suspended sediment concentrations in the Mekong River Basin via semi-supervised-based deep neural networks

学校读者我要写书评

暂无评论

science of the Total Environment 2024年 955卷 176758页

作者： Nguyen, Thi Thu Ha Vu, Duc Quang Doan, Ngoc Phu Chi, Huynh Thi Khanh Li, Peixin Binh, Doan Van An, Yimeng Dung, Pham Tuan Hoang, Tuan A. Son, Mai Thai Laboratory of Environmental Sciences and Climate Change Institute for Computational Science and Artificial Intelligence Van Lang University Ho Chi Minh City Viet Nam Faculty of Environment School of Technology Van Lang University Ho Chi Minh City Viet Nam Department of Computer Science and Information System Thai Nguyen University of Education Viet Nam School of Electronics Electrical Engineering and Computer Science Queen's University Belfast United Kingdom College of Economics and Management Nanjing University of Aeronautics and Astronautics China Master program in Water Technology Reuse and Management Faculty of Engineering Vietnamese-German University Binh Duong Province Ben Cat City Viet Nam

The Mekong River Basin (MRB) is crucial for the livelihoods of over 60 million people across six Southeast Asian countries. Understanding long-term sediment changes is crucial for management and contingency plans, but the sediment concentration data in the MRB are extremely sporadic, making analysis challenging. This study focuses on reconstructing long-term suspended sediment concentration (SSC) data using a novel semi-supervised machine learning (ML) model. The key idea of this approach is to exploit abundant available hydroclimate data to reduce training overfitting rather than solely relying on sediment concentration data, thus enhancing the accuracy of the employed ML models. Extensive experiments on daily hydroclimate and SSC data obtained from 1979 to 2019 at the three main stations (i.e., Chiang Saen, Nong Khai, and Mukdahan) are conducted to demonstrate the superior performance of the proposed method compared to the state-of-the-art supervised techniques (i.e., Random Forest, XGBoost, CatBoost, MLP, CNN, and LSTM), and surpasses existing semi-supervised methods (i.e., CoReg, ⊓ Model, ICT, and Mean Teacher). This approach is the first semi-supervised method to reconstruct sediment data in the field and has the potential for broader application in other river systems. © 2024 The Author(s)

关键词： Self-supervised learning

Local order metrics for many-particle systems across length scales

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Maher, Charles Emmett Torquato, Salvatore Department of Chemistry Princeton University New Jersey08544 United States Department of Chemistry Princeton University PrincetonNJ08544 United States Department of Physics Princeton University PrincetonNJ08544 United States Princeton Institute for the Science and Technology of Materials Princeton University PrincetonNJ08544 United States Program in Applied and Computational Mathematics Princeton University PrincetonNJ08544 United States

Formulating order metrics that sensitively quantify the degree of order/disorder in many-particle systems in d-dimensional Euclidean space Rd across length scales is an outstanding challenge in physics, chemistry, and materials science. Since an infinite set of n-particle correlation functions is required to fully characterize a system, one must settle for a reduced set of structural information, in practice. We initiate a program to use the local number variance σN2 (R) associated with a spherical sampling window of radius R (which encodes pair correlations) and an integral measure derived from it ΣN(Ri, Rj) that depends on two specified radial distances Ri and Rj. Across the first three space dimensions (d = 1, 2, 3), we find these metrics can sensitively describe and categorize the degree of order/disorder of 41 different models of antihyperuniform, nonhyperuniform, disordered hyperuniform, and ordered hyperuniform many-particle systems at a specified length scale R. Using our local variance metrics, we demonstrate the importance of assessing order/disorder with respect to a specific value of R. These local order metrics could also aid in the inverse design of structures with prescribed length-scale-specific degrees of order/disorder that yield desired physical properties. In future work, it would be fruitful to explore the use of higher-order moments of the number of points within a spherical window of radius R [S. Torquato et al., Phys. Rev. X, 11, 021028 (2021)] to devise even more sensitive order metrics. © 2024, CC BY.

关键词： Inverse problems

Bridging scales with Machine Learning: From first principles statistical mechanics to continuum phase field computations to study order-disorder transitions in LixCoO2

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Teichert, G.H. Das, S. Shojaei, M. Faghih Holber, J. Mueller, T. Hung, L. Gavini, V. Garikipati, K. Department of Mechanical Engineering University of Michigan United States Toyota Research Institute Los AltosCA United States Department of Materials Science & Engineering University of Michigan United States Department of Mathematics University of Michigan United States Michigan Institute for Computational Discovery & Engineering University of Michigan United States Applied Physics Program University of Michigan United States Materials Science & Engineering Johns Hopkins University United States

LixTMO2 (TM=Ni, Co, Mn) forms an important family of cathode materials for Li-ion batteries, whose performance is strongly governed by Li composition-dependent crystal structure and phase stability. Here, we use LixCoO2 (LCO) as a model system to benchmark a machine learning-enabled framework for bridging scales in materials physics. We focus on two scales: (a) assemblies of thousands of atoms described by density functional theory-informed statistical mechanics, and (b) continuum phase field models to study the dynamics of order-disorder transitions in LCO. Central to the scale bridging is the rigorous, quantitatively accurate, representation of the free energy density and chemical potentials of this material system by coarsegraining formation energies for specific atomic configurations. We develop active learning workflows to train recently developed integrable deep neural networks for such high-dimensional free energy density and chemical potential functions. The resulting, first principles-informed, machine learning-enabled, phase-field computations allow us to study LCO cathodes’ phase evolution in terms of temperature, morphology, charge cycling and particle size. © 2023, CC BY-NC-SA.

关键词： Cathodes

Theoretical Insights into Co2 Electroreduction on Single and Dual Heteroatom-Doped Diamonds

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Injongkol, Yuwanda Zhang, Rui-Qin Montoya, Alejandro Rungrotmongkol, Thanyada Jungsuttiwong, Siriporn Center of Excellence in Biocatalyst and Sustainable Biotechnology Department of Biochemistry Faculty of Science Chulalongkorn University Bangkok10330 Thailand City University of Hong Kong Hong Kong School of Chemical and Biomolecular Engineering The University of Sydney SydneyNSW2006 Australia Program in Bioinformatics and Computational Biology Graduate School Chulalongkorn University Bangkok10330 Thailand Department of Chemistry Faculty of Science Ubon Ratchathani University Ubon Ratchathani34190 Thailand

We employed Periodic Density Functional Theory (DFT) to investigate the catalytic activity of single and double dopant of boron (B), nitrogen (N), and phosphorous (P) anchored on diamond surfaces for the CO2 reduction reaction (CO2RR) by analyzing reaction energy profiles. Our findings show that the double-doped catalyst exhibits lower energy barriers in CO2RR compared to the single dopants. Specifically, when considering BB and NN configurations, there is a thermodynamic preference towards the formation of formic acid (HCOOH) with overpotentials of 0.40 and 0.09 V, respectively. The co-doped catalyst comprising B and N (BN) demonstrates a tendency towards the formation of HCOOH without requiring any applied overpotentials. Notably, BN outperforms other catalysts, occupying the top position on the volcano plot, indicating the lowest limiting potential (UL), remarkable thermal stability, and the ability to suppress the competing hydrogen evolution reaction (HER). This research provides valuable insights into the product differentiation in the electroreduction of CO2. © 2023, The Authors. All rights reserved.

关键词： Density functional theory

Structural properties of hyperuniform Voronoi networks

学校读者我要写书评

暂无评论

Physical Review E 2025年第3期111卷 034123-034123页

作者： Eli Newby Wenlong Shi Yang Jiao Reka Albert Salvatore Torquato Department of Physics Pennsylvania State University University Park Pennsylvania 16802 USA Materials Science and Engineering Arizona State University Tempe Arizona 85287 USA Department of Physics Arizona State University Tempe Arizona 85287 USA Department of Chemistry Princeton University Princeton New Jersey 08544 USA Department of Physics Princeton University Princeton New Jersey 08544 USA Princeton Institute of Materials Princeton University Princeton New Jersey 08544 USA Program in Applied and Computational Mathematics Princeton University Princeton New Jersey 08544 USA

Disordered hyperuniform many-particle systems are recently discovered exotic states of matter, characterized by the complete suppression of normalized infinite-wavelength density fluctuations, as in perfect crystals, while lacking conventional long-range order, as in liquids and glasses. In this work, we begin a program to quantify the structural properties of nonhyperuniform and hyperuniform networks. In particular, large two-dimensional (2D) Voronoi networks (graphs) containing approximately 10,000 nodes are created from a variety of different point configurations, including the antihyperuniform hyperplane intersection process (HIP), nonhyperuniform Poisson process, nonhyperuniform random sequential addition (RSA) saturated packing, and both non-stealthy and stealthy hyperuniform point processes. We carry out an extensive study of the Voronoi-cell area distribution of each of the networks by determining multiple metrics that characterize the distribution, including their average areas and corresponding variances as well as higher-order cumulants (i.e., skewness γ1 and excess kurtosis γ2). We show that the HIP distribution is far from Gaussian, as evidenced by a high skewness (γ1=3.16) and large positive excess kurtosis (γ2=16.2). The Poisson (with γ1=1.07 and γ2=1.79) and non-stealthy hyperuniform (with γ1=0.257 and γ2=0.0217) distributions are Gaussian-like distributions, since they exhibit a small but positive skewness and excess kurtosis. The RSA (with γ1=0.450 and γ2=−0.0384) and the highest stealthy hyperuniform distributions (with γ1=0.0272 and γ2=−0.0626) are also non-Gaussian because of their low skewness and negative excess kurtosis, which is diametrically opposite of the non-Gaussian behavior of the HIP. The fact that the cell-area distributions of large, finite-sized RSA and stealthy hyperuniform networks (e.g., with N≈10,000 nodes) are narrower, have larger peaks, and smaller tails than a Gaussian distribution implies that in the thermodynamic limit th

关键词： Gaussian distribution

Ordered and disordered stealthy hyperuniform point patterns across spatial dimensions

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Morse, Peter K. Steinhardt, Paul J. Torquato, Salvatore Department of Chemistry Princeton University PrincetonNJ08544 United States Department of Physics Princeton University PrincetonNJ08544 United States Princeton Institute of Materials Princeton University PrincetonNJ08544 United States Princeton Center for Theoretical Science Princeton University PrincetonNJ08544 United States Program in Applied and Computational Mathematics Princeton University PrincetonNJ08544 United States

In previous work [Phys. Rev. X 5, 021020 (2015)], it was shown that stealthy hyperuniform systems can be regarded as hard spheres in Fourier-space in the sense that the the structure factor is exactly zero in a spherical region around the origin in analogy with the pair-correlation function of real-space hard spheres. In this work, we exploit this correspondence to confirm that the densest Fourier-space hard-sphere system is that of a Bravais lattice. This is in contrast to real-space hard-spheres, whose densest configuration is conjectured to be disordered. We also extend the virial series previously suggested for disordered stealthy hyperuniform systems to higher dimensions in order to predict spatial decorrelation as function of dimension. This prediction is then borne out by numerical simulations of disordered stealthy hyperuniform ground states in dimensions d = 2-8. © 2024, CC BY.

关键词： Ground state

Hyperuniformity of maximally random jammed packings of hyperspheres across spatial dimensions

学校读者我要写书评

暂无评论

Physical Review E 2023年第6期108卷 064602-064602页

作者： Charles Emmett Maher Yang Jiao Salvatore Torquato Department of Chemistry Princeton University Princeton New Jersey 08544 USA Materials Science and Engineering Arizona State University Tempe Arizona 85287 USA Department of Physics Arizona State University Tempe Arizona 85287 USA Department of Physics Princeton University Princeton New Jersey 08544 USA Princeton Materials Institute Princeton University Princeton New Jersey 08544 USA Program in Applied and Computational Mathematics Princeton University Princeton New Jersey 08544 USA

The maximally random jammed (MRJ) state is the most random (i.e., disordered) configuration of strictly jammed (mechanically rigid) nonoverlapping objects. MRJ packings are hyperuniform, meaning their long-wavelength density fluctuations are anomalously suppressed compared to typical disordered systems, i.e., their structure factors S(k) tend to zero as the wave number |k| tends to zero. Here we show that generating high-quality strictly jammed states for Euclidean space dimensions d=3,4, and 5 is of paramount importance in ensuring hyperuniformity and extracting precise values of the hyperuniformity exponent α>0 for MRJ states, defined by the power-law behavior of S(k)∼|k|α in the limit |k|→0. Moreover, we show that for fixed d it is more difficult to ensure jamming as the particle number N increases, which results in packings that are nonhyperuniform. Free-volume theory arguments suggest that the ideal MRJ state does not contain rattlers, which act as defects in numerically generated packings. As d increases, we find that the fraction of rattlers decreases substantially. Our analysis of the largest truly jammed packings suggests that the ideal MRJ packings for all dimensions d≥3 are hyperuniform with α=d−2, implying the packings become more hyperuniform as d increases. The differences in α between MRJ packings and the recently proposed Manna-class random close packed (RCP) states, which were reported to have α=0.25 in d=3 and be nonhyperuniform (α=0) for d=4 and d=5, demonstrate the vivid distinctions between the large-scale structure of RCP and MRJ states in these dimensions. Our paper clarifies the importance of the link between true jamming and hyperuniformity and motivates the development of an algorithm to produce rattler-free three-dimensional MRJ packings.

关键词： Granular solids