检索结果-内蒙古大学图书馆

Foundations of symbolic languages for model interpretability

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Arenas, Marcelo Baez, Daniel Barceló, Pablo Pérez, Jorge Subercaseaux Roa, Bernardo Anibal Department of Computer Science PUC-Chile Chile Institute for Mathematical and Computational Engineering PUC-Chile Chile Department of Computer Science Universidad de Chile Chile Millennium Institute for Foundational Research on Data Chile Carnegie Mellon University United States

Several queries and scores have recently been proposed to explain individual predictions over ML models. Given the need for flexible, reliable, and easy-to-apply interpretability methods for ML models, we foresee the need for developing declarative languages to naturally specify different explainability queries. We do this in a principled way by rooting such a language in a logic, called FOIL, that allows for expressing many simple but important explainability queries, and might serve as a core for more expressive interpretability languages. We study the computational complexity of FOIL queries over two classes of ML models often deemed to be easily interpretable: decision trees and OBDDs. Since the number of possible inputs for an ML model is exponential in its dimension, the tractability of the FOIL evaluation problem is delicate but can be achieved by either restricting the structure of the models or the fragment of FOIL being evaluated. We also present a prototype implementation of FOIL wrapped in a high-level declarative language and perform experiments showing that such a language can be used in practice. © 2021, CC BY.

关键词： Decision trees

Characterizing the SLOPE Trade-off: A Variational Perspective and the Donoho-Tanner Limit

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Bu, Zhiqi Klusowski, Jason M. Rush, Cynthia Su, Weijie J. Graduate Group in Applied Mathematics and Computational Science University of Pennsylvania United States Department of Operations Research and Financial Engineering Princeton University United States Department of Statistics Columbia University United States Department of Statistics and Data Science University of Pennsylvania United States

Sorted l1 regularization has been incorporated into many methods for solving high-dimensional statistical estimation problems, including the SLOPE estimator in linear regression. In this paper, we study how this relatively new regularization technique improves variable selection by characterizing the optimal SLOPE trade-off between the false discovery proportion (FDP) and true positive proportion (TPP) or, equivalently, between measures of type I error and power. Assuming a regime of linear sparsity and working under Gaussian random designs, we obtain an upper bound on the optimal trade-off for SLOPE, showing its capability of breaking the Donoho-Tanner power limit. To put it into perspective, this limit is the highest possible power that the Lasso, which is perhaps the most popular l1-based method, can achieve even with arbitrarily strong effect sizes. Next, we derive a tight lower bound that delineates the fundamental limit of sorted l1 regularization in optimally trading the FDP off for the TPP. Finally, we show that on any problem instance, SLOPE with a certain regularization sequence outperforms the Lasso, in the sense of having a smaller FDP, larger TPP and smaller l2 estimation risk simultaneously. Our proofs are based on a novel technique that reduces a calculus of variations problem to a class of infinite-dimensional convex optimization problems and a very recent result from approximate message passing theory. Copyright © 2021, The Authors. All rights reserved.

关键词： Message passing

End-to-end AI Framework for Interpretable Prediction of Molecular and Crystal Properties

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Park, Hyun Zhu, Ruijie Huerta, E.A. Chaudhuri, Santanu Tajkhorshid, Emad Cooper, Donny Theoretical and Computational Biophysics Group Beckman Institute for Advanced Science and Technology University of Illinois at Urbana-Champaign UrbanaIL61801 United States Department of Materials Science and Engineering Northwestern University EvanstonIL60208 United States Data Science and Learning Division Argonne National Laboratory LemontIL60439 United States Department of Computer Science The University of Chicago ChicagoIL60637 United States Department of Physics University of Illinois at Urbana-Champaign UrbanaIL61801 United States Multiscale Materials and Manufacturing Lab University of Illinois Chicago ChicagoIL60607 United States Department of Biochemistry University of Illinois at Urbana-Champaign UrbanaIL61801 United States Center for Biophysics and Quantitative Biology University of Illinois at Urbana-Champaign UrbanaIL61801 United States Computational Science and Engineering Data Science and AI Department TotalEnergies EP Research & Technology USA LLC HoustonTX77002 United States

We introduce an end-to-end computational framework that allows for hyperparameter optimization using the DeepHyper library, accelerated model training, and interpretable AI inference. The framework is based on state-of-the-art AI models including CGCNN, PhysNet, SchNet, MPNN, MPNN-transformer, and TorchMD-NET. We employ these AI models along with the benchmark QM9, hMOF, and MD17 datasets to showcase how the models can predict user-specified material properties within modern computing environments. We demonstrate transferable applications in the modeling of small molecules, inorganic crystals and nanoporous metal organic frameworks with a unified, standalone framework. We have deployed and tested this framework in the ThetaGPU supercomputer at the Argonne Leadership Computing Facility, and in the Delta supercomputer at the National Center for Supercomputing Applications to provide researchers with modern tools to conduct accelerated AI-driven discovery in leadership-class computing environments. We release these digital assets as open source scientific software in GitLab, and ready-to-use Jupyter notebooks in Google Colab. Copyright © 2022, The Authors. All rights reserved.

关键词： Organometallics

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

学校读者我要写书评

暂无评论

arXiv 2024年

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. © 2024, CC BY-NC-ND.

关键词： Large datasets

Temporal networks provide a unifying understanding of the evolution of cooperation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Li, Aming Meng, Yao Zhou, Lei Masuda, Naoki Wang, Long Center for Systems and Control College of Engineering Peking University Beijing100871 China Center for Multi-Agent Research Institute for Artificial Intelligence Peking University Beijing100871 China School of Automation Beijing Institute of Technology Beijing100081 China Department of Mathematics State University of New York at Buffalo BuffaloNY14260 United States Computational and Data-Enabled Science and Engineering Program State University of New York at Buffalo BuffaloNY14260 United States

Understanding the evolution of cooperation in structured populations represented by networks is a problem of long research interest, and a most fundamental and widespread property of social networks related to cooperation phenomena is that the node’s degree (i.e., number of edges connected to the node) is heterogeneously distributed. Previous results indicate that static heterogeneous (i.e., degree-heterogeneous) networks promote cooperation in stationarity compared to static regular (i.e., degree-homogeneous) networks if equilibrium dynamics starting from many cooperators and defectors is employed. However, the above conclusion reverses if we employ non-equilibrium stochastic processes to measure the fixation probability for cooperation, i.e., the probability that a single cooperator successfully invades a population. Here we resolve this conundrum by analyzing the fixation of cooperation on temporal (i.e., time-varying) networks. We theoretically prove and numerically confirm that on both synthetic and empirical networks, contrary to the case of static networks, temporal heterogeneous networks can promote cooperation more than temporal regular networks in terms of the fixation probability of cooperation. Given that the same conclusion is known for the equilibrium fraction of cooperators on temporal networks, the present results provide a unified understanding of the effect of temporal degree heterogeneity on promoting cooperation across two main analytical frameworks, i.e., equilibrium and non-equilibrium ones. Copyright © 2023, The Authors. All rights reserved.

关键词： Heterogeneous networks

Quasi-van der Waals Epitaxial Growth of γ'-GaSe Nanometer-Thick Films on GaAs(111)B Substrates

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yu, Mingyu Iddawela, Sahani Amaya Wang, Jiayang Hilse, Maria Thompson, Jessica L. Hickey, Danielle Reifsnyder Sinnott, Susan B. Law, Stephanie Department of Materials Science and Engineering University of Delaware 201 Dupont Hall 127 The Green NewarkDE19716 United States Department of Chemistry The Pennsylvania State University University ParkPA16802 United States Department of Materials Science and Engineering The Pennsylvania State University University ParkPA16802 United States Materials Research Institute The Pennsylvania State University University ParkPA16802 United States 2D Crystal Consortium Materials Innovation Platform The Pennsylvania State University University ParkPA16802 United States Institute for Computational and Data Science The Pennsylvania State University University ParkPA16802 United States Penn State Institute of Energy and the Environment The Pennsylvania State University University ParkPA16802 United States

GaSe is an important member of the post-transition metal chalcogenide family and is an emerging two-dimensional (2D) semiconductor material. Because it is a van der Waals material, it can be fabricated into atomic-scale ultrathin films, making it suitable for the preparation of compact, heterostructure devices. In addition, GaSe possesses unusual optical and electronic properties, such as a shift from an indirect-bandgap single-layer film to a direct-bandgap bulk material, rare intrinsic p-type conduction, and nonlinear optical behaviors. These properties make GaSe an appealing candidate for the fabrication of field-effect transistors, photodetectors, and photovoltaics. However, the wafer-scale production of pure GaSe single crystal thin films remains challenging. This study develops an approach for the direct growth of nanometer-thick GaSe films on GaAs substrates using molecular beam epitaxy. It yields smooth thin GaSe films with the rare γ'-polymorph. We analyze the formation mechanism of γ'-GaSe using density functional theory and speculate that it is stabilized by Ga vacancies since the formation enthalpy of γ'-GaSe tends to become lower than that of other polymorphs when the Ga vacancy concentration increases. Finally, we investigate the growth conditions of GaSe, providing valuable insights for exploring 2D/3D quasi-van der Waals epitaxial growth. © 2024, CC BY-NC-ND.

关键词： Electronic properties

Opinion dynamics on tie-decay networks

学校读者我要写书评

暂无评论

Physical Review Research 2021年第2期3卷 023249-023249页

作者： Kashin Sugishita Mason A. Porter Mariano Beguerisse-Díaz Naoki Masuda Department of Mathematics State University of New York at Buffalo Buffalo New York 14260-2900 USA Computational and Data-Enabled Science and Engineering Program State University of New York at Buffalo Buffalo New York 14260-5030 USA and Faculty of Science and Engineering Waseda University 169-8555 Tokyo Japan

In social networks, interaction patterns typically change over time. We study opinion dynamics on tie-decay networks in which tie strength increases instantaneously when there is an interaction and decays exponentially between interactions. Specifically, we formulate continuous-time Laplacian dynamics and a discrete-time DeGroot model of opinion dynamics on these tie-decay networks, and we carry out numerical computations for the continuous-time Laplacian dynamics. We examine the speed of convergence by studying the spectral gaps of combinatorial Laplacian matrices of tie-decay networks. First, we compare the spectral gaps of the Laplacian matrices of tie-decay networks that we construct from empirical data with the spectral gaps for corresponding randomized and aggregate networks. We find that the spectral gaps for the empirical networks tend to be smaller than those for the randomized and aggregate networks. Second, we study the spectral gap as a function of the tie-decay rate and time. Intuitively, we expect small tie-decay rates to lead to fast convergence because the influence of each interaction between two nodes lasts longer for smaller decay rates. Moreover, as time progresses and more interactions occur, we expect eventual convergence. However, we demonstrate that the spectral gap need not decrease monotonically with respect to the decay rate or increase monotonically with respect to time. Our results highlight the importance of the interplay between the times that edges strengthen and decay in temporal networks.

关键词： Network structure Social networks Consensus & non-consensus models

Multi-Objective Latent Space Optimization of Generative Molecular Design Models

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Nafiz Abeer, A.N.M. Urban, Nathan M. Ryan Weil, M. Alexander, Francis J. Yoon, Byung-Jun Department of Electrical and Computer Engineering Texas A&M University College StationTX77843 United States Computational Science Initiative Brookhaven National Laboratory UptonNY11973 United States Strategic and Data Science Initiatives Frederick National Laboratory FrederickMD21702 United States Argonne National Laboratory LemontIL60439 United States

Molecular design based on generative models, such as variational autoencoders (VAEs), has become increasingly popular in recent years due to its efficiency for exploring high-dimensional molecular space to identify molecules with desired properties. While the efficacy of the initial model strongly depends on the training data, the sampling efficiency of the model for suggesting novel molecules with enhanced properties can be further enhanced via latent space optimization. In this paper, we propose a multi-objective latent space optimization (LSO) method that can significantly enhance the performance of generative molecular design (GMD). The proposed method adopts an iterative weighted retraining approach, where the respective weights of the molecules in the training data are determined by their Pareto efficiency. We demonstrate that our multi-objective GMD LSO method can significantly improve the performance of GMD for jointly optimizing multiple molecular properties. © 2022, CC BY-NC-SA.

关键词： Molecular docking

Effects of concurrency on epidemic spreading in Markovian temporal networks

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liu, Ruodan Ogura, Masaki Reis, Elohim Fonseca Dos Masuda, Naoki Department of Mathematics State University of New York at Buffalo BuffaloNY14260-2900 United States Graduate School of Information Science and Technology Osaka University Osaka Suita565-0871 Japan Computational and Data-Enabled Sciences and Engineering Program State University of New York at Buffalo BuffaloNY14260-5030 United States Faculty of Science and Engineering Waseda University Tokyo169-8555 Japan

The concurrency of edges, quantified by the number of edges that share a common node at a given time point, may be an important determinant of epidemic processes in temporal networks. We propose theoretically tractable Markovian temporal network models in which each edge flips between the active and inactive states in continuous time. The different models have different amounts of concurrency while we can tune the models to share the same statistics of edge activation and deactivation (and hence the fraction of time for which each edge is active) and the structure of the aggregate (i.e., static) network. We analytically calculate the amount of concurrency of edges sharing a node for each model. We then numerically study effects of concurrency on epidemic spreading in the stochastic susceptible-infectious-susceptible and susceptible-infectious-recovered dynamics on the proposed temporal network models. We find that the concurrency enhances epidemic spreading near the epidemic threshold while this effect is small in many cases. Furthermore, when the infection rate is substantially larger than the epidemic threshold, the concurrency suppresses epidemic spreading in a majority of cases. In sum, our numerical simulations suggest that the impact of concurrency on enhancing epidemic spreading within our model is consistently present near the epidemic threshold but modest. The proposed temporal network models are expected to be useful for investigating effects of concurrency on various collective dynamics on networks including both infectious and other dynamics. Copyright © 2022, The Authors. All rights reserved.

关键词： Dynamics