检索结果-内蒙古大学图书馆

mFI-PSO: A Flexible and Effective Method in Adversarial Image Generation for Deep Neural Networks

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Shu, Hai Shi, Ronghua Jia, Qiran Zhu, Hongtu Chen, Ziqi Department of Biostatistics School of Global Public Health New York University New YorkNY United States School of Mathematics and Statistics Central South University Changsha China Departments of Biostatistics Statistics Computer Science and Genetics The University of North Carolina at Chapel Hill Chapel HillNC United States Key Laboratory of Advanced Theory and Application in Statistics and Data Science-MOE School of Statistics East China Normal University Shanghai China

Deep neural networks (DNNs) have achieved great success in image classification, but can be very vulnerable to adversarial attacks with small perturbations to images. To improve adversarial image generation for DNNs, we develop a novel method, called mFI-PSO, which utilizes a Manifoldbased First-order Influence measure for vulnerable image and pixel selection and the Particle Swarm Optimization for various objective functions. Our mFI-PSO can thus effectively design adversarial images with flexible, customized options on the number of perturbed pixels, the misclassification probability, and the targeted incorrect class. Experiments demonstrate the flexibility and effectiveness of our mFI-PSO in adversarial attacks and its appealing advantages over some popular methods. Copyright © 2020, The Authors. All rights reserved.

关键词： Particle swarm optimization (PSO)

Optimal subsampling for the Cox proportional hazards model with massive survival data

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Qiao, Nan Li, Wangcheng Xiao, Feng Lin, Cunjie Zhou, Yong Center for Applied Statistics Renmin University of China Beijing100872 China School of Statistics Renmin University of China Beijing100872 China School of Statistics Beijing Normal University Beijing100875 China Sichuan Rural Credit Union Chengdu61000 China Key Laboratory of Advanced Theory and Application in Statistics and Data Science-MOE Shanghai200062 China Academy of Statistics and Interdisciplinary Sciences East China Normal University Shanghai200062 China

The use of massive survival data has become common in survival analysis. In this study, a subsampling algorithm is proposed for the Cox proportional hazards model with time-dependent covariates when the sample is extraordinarily large but computing resources are relatively limited. A subsample estimator is developed by maximizing the weighted partial likelihood;it is shown to have consistency and asymptotic normality. By minimizing the asymptotic mean squared error of the subsample estimator, the optimal subsampling probabilities are formulated with explicit expressions. Simulation studies show that the proposed method can satisfactorily approximate the estimator of the full dataset. The proposed method is then applied to corporate loan and breast cancer datasets, with different censoring rates, and the outcomes confirm its practical advantages. Copyright © 2023, The Authors. All rights reserved.

关键词： Mean square error

Fractional stochastic wave equation driven by a Gaussian noise rough in space

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Song, Jian Song, Xiaoming Xu, Fangjun School of Mathematics Shandong University Jinan Shandong250100 China Department of Mathematics Drexel University PhiladelphiaPA19104 United States Key Laboratory of Advanced Theory and Application in Statistics and Data Science - MOE School of Statistics East China Normal University Shanghai200062 China NYU-ECNU Institute of Mathematical Sciences NYU Shanghai Shanghai200062 China

In this article, we consider fractional stochastic wave equations on R driven by a multiplicative Gaussian noise which is white/colored in time and has the covariance of a fractional Brownian motion with Hurst parameter H ∈ (1/4, 1/2) in space. We prove the existence and uniqueness of the mild Skorohod solution, establish lower and upper bounds for the p-th moment of the solution for all p ≥ 2, and obtain the Hölder continuity in time and space variables for the *** Codes 60H07, 60H15, 60G15 Copyright © 2019, The Authors. All rights reserved.

关键词： Gaussian noise (electronic)

On the estimation of high-dimensional integrated covariance matrices based on high-frequency data with multiple transactions

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Wang, Moming Xia, Ningning Zhou, Yong School of Statistics and Management Shanghai University of Finance and Economics 777 Guo Ding Road Shanghai200433 China School of Statistics and Management Shanghai Key Laboratory of Financial Information Technology Shanghai University of Finance and Economics 777 Guo Ding Road Shanghai200433 China Faculty of Economics and Management East China Normal University Shanghai200241 China Academy of Mathematics and Systems Science Chinese Academy of Science Beijing100190 China Key Laboratory of Advanced Theory and Application in Statistics and Data Science MOE Academy of Statistics and Interdisciplinary Sciences School of Statistics East China Normal University Shanghai200062 China

High-frequency data in financial markets often include multiple transactions at each recording time due to the mechanism of recording. Using random matrix theory, this paper considers the estimation of integrated covariance (ICV) matrices of high-dimensional diffusion processes based on multiple high-frequency observations. We start by studying the estimator, the time-variation adjusted realized covariance (TVA) matrix proposed by Zheng and Li (2011), without microstructure noise. We show that in the high-dimensional case, for a class C of diffusion processes, the limiting spectral distribution (LSD) of the averaged TVA depends not only on that of the ICV but also on the number of multiple transactions at each recording time. However, in practice, observed prices are always contaminated by market microstructure noise. Thus, we also study the limiting behavior of pre-averaging averaged TVA matrices based on noisy multiple observations. We show that for processes in class C, the pre-averaging averaged TVA has two desirable properties: it eliminates the effects of microstructure noise and multiple transactions, and its LSD depends solely on that of the ICV matrix. Further, three types of nonlinear shrinkage estimators of the ICV matrix are proposed based on high-frequency noisy multiple observations. Simulation studies support our theoretical results and demonstrate the finite sample performance of the proposed estimators. Finally, high-frequency portfolio strategies are evaluated under these estimators in real data analysis. Copyright © 2019, The Authors. All rights reserved.

关键词： Covariance matrix

Class-Imbalanced Semi-Supervised Learning for Large-Scale Point Cloud Semantic Segmentation Via Decoupling Optimization

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Li, Mengtian Lin, Shaohui Wang, Zihan Shen, Yunhang Zhang, Baochang Ma, Lizhuang Visual Media Intelligent Processing Research Group Department of Film and Television Engineering Shanghai University Shanghai200072 China School of Computer Science and Technology East China Normal University Shanghai200062 China Key Laboratory of Advanced Theory and Application in Statistics and Data Science Ministry of Education China School of Automation Science and Electrical Engineering Beihang University Beijing100191 China Youtu Lab Tencent China

Semi-supervised learning (SSL), thanks to the significant reduction of data annotation costs, has been an active research topic for large-scale 3D scene understanding. However, the existing SSL-based methods suffer from severe training bias, mainly due to class imbalance and long-tail distributions of the point cloud data. As a result, they lead to a biased prediction for the tail class segmentation. In this paper, we introduce a new decoupling optimization framework, which disentangles feature representation learning and classifier in an alternative optimization manner to shift the bias decision boundary effectively. In particular, we first employ two-round pseudo-label generation to select unlabeled points across head-to-tail classes. We further introduce multi-class imbalanced focus loss to adaptively pay more attention to feature learning across head-to-tail classes. We fix the backbone parameters after feature learning and retrain the classifier using ground-truth points to update its parameters. Ex-✩Fully documented templates are available in the elsarticle package on CTAN. © 2023, The Authors. All rights reserved.

关键词： Semantic Segmentation

Filter Pruning for Efficient CNNs via Knowledge-driven Differential Filter Sampler

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lin, Shaohui Huang, Wenxuan Xie, Jiao Zhang, Baochang Shen, Yunhang Yu, Zhou Han, Jungong Doermann, David The School of Computer Science and Technology East China Normal University Shanghai China The Key Laboratory of Advanced Theory and Application in Statistics and Data Science Ministry of Education China The Department of Automation School of Aerospace Engineering Xiamen University Xiamen China The Institute of Artificial Intelligence Beihang University Beijing China Zhongguancun Laboratory Beijing China The Youtu Lab Tencent Shanghai China The School of Statistics East China Normal University Shanghai China The Department of Computer Science University of Sheffield United Kingdom The University at Buffalo BuffaloNY United States

Filter pruning simultaneously accelerates the computation and reduces the memory overhead of CNNs, which can be effectively applied to edge devices and cloud services. In this paper, we propose a novel Knowledge-driven Differential Filter Sampler (KDFS) with Masked Filter Modeling (MFM) framework for filter pruning, which globally prunes the redundant filters based on the prior knowledge of a pre-trained model in a differential and non-alternative optimization. Specifically, we design a differential sampler with learnable sampling parameters to build a binary mask vector for each layer, determining whether the corresponding filters are redundant. To learn the mask, we introduce masked filter modeling to construct PCA-like knowledge by aligning the intermediate features from the pre-trained teacher model and the outputs of the student decoder taking sampling features as the input. The mask and sampler are directly optimized by the Gumbel-Softmax Straight-Through Gradient Estimator in an end-to-end manner in combination with global pruning constraint, MFM reconstruction error, and dark knowledge. Extensive experiments demonstrate the proposed KDFS’s effectiveness in compressing the base models on various datasets. For instance, the pruned ResNet-50 on ImageNet achieves 55.36% computation reduction, and 42.86% parameter reduction, while only dropping 0.35% Top-1 accuracy, significantly outperforming the state-of-the-art methods. The code is available at https://***/Osilly/KDFS. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning

On stochastic control problems with higher-order moments

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Wang, Yike Liu, Jingzhen Bensoussan, Alain Yiu, Ka-Fai Cedric Wei, Jiaqin School of Finance Chongqing Technology and Business University Chongqing400067 China China Institute for Actuarial Science Central University of Finance and Economics Beijing100081 China Naveen Jindal School of Management University of Texas at Dallas DallasTX75083 United States Department of Applied Mathematics The Hong Kong Polytechnic University Kowloon Hunghom Hong Kong Key Laboratory of Advanced Theory and Application in Statistics and Data Science MOE School of Statistics East China Normal University Shanghai200241 China

In this paper, we focus on a class of time-inconsistent stochastic control problems, where the objective function includes the mean and several higher-order central moments of the terminal value of state. To tackle the time-inconsistency, we seek both the closed-loop and the open-loop Nash equilibrium controls as time-consistent solutions. We establish a partial differential equation (PDE) system for deriving a closed-loop Nash equilibrium control, which does not include the equilibrium value function and is different from the extended Hamilton-Jacobi-Bellman (HJB) equations as in Björk et al. (Finance Stoch. 21: 331-360, 2017). We show that our PDE system is equivalent to the extended HJB equations that seems difficult to be solved for our higher-order moment problems. In deriving an open-loop Nash equilibrium control, due to the non-separable higher-order moments in the objective function, we make some moment estimates in addition to the standard perturbation argument for developing a maximum principle. Then, the problem is reduced to solving a flow of forward-backward stochastic differential equations. In particular, we investigate linear controlled dynamics and some objective functions affine in the mean. The closed-loop and the open-loop Nash equilibrium controls are identical, which are independent of the state value, random path and the preference on the odd-order central moments. By sending the highest order of central moments to infinity, we obtain the time-consistent solutions to some control problems whose objective functions include some penalty functions for *** Codes Primary: 93E20, 91G80, Secondary: 91B08, 49N90 Copyright © 2024, The Authors. All rights reserved.

关键词： Nash equilibrium

Native spaces and generalization of Wu functions

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huang, Yixuan Wu, Zongmin Zhu, Shengxin School of Mathematics Beijing Normal University China Laboratory of Mathematics and Complex Systems Ministry of Education China Shanghai Key Laboratory for Contemporary Applied Mathematics School of Mathematical Sciences Fudan University Shanghai China Shanghai Center for Mathematical Sciences Shanghai China School of Big Data and Statistics Anhui University Hefei China Research Center for Mathematics Advanced Institute of Natural Science Beijing Normal University Zhuhai519087 China Guangdong Provincial Key Laboratory of Interdisciplinary Research and Application for Data Science BNU-HKBU United International College Zhuhai519087 China

We prove that the native space of a Wu function is a dense subspace of a Sobolev space. An explicit characterization of the native spaces of Wu functions is given. Three definitions of Wu functions are introduced and proven to be equivalent. Based on these new equivalent definitions and the so called f-form tricks, we can generalize the Wu functions into the even-dimensional spaces R2k, while the original Wu functions are only defined in the odd-dimensional spaces R2k+1. Such functions in even-dimensional spaces are referred to as the ‘missing Wu functions’. Furthermore we can generalize the Wu functions into ‘fractional’-dimensional spaces. We call all these Wu functions the generalized Wu functions. The closed form of the generalized Wu functions are given in terms of hypergeometric functions. Finally we prove that the Wu functions and the missing Wu functions can be written as linear combinations of the generalized Wendland *** Codes 65J05, 33C90, 41A05, 41A30, 41A65, 41A63 Copyright © 2023, The Authors. All rights reserved.

关键词： Sobolev spaces

Dependency, Reciprocity, and Mentorship in Long-Term Research Collaboration: A Co-Authorship Matrix-Based Multivariate Time Series Analysis

学校读者我要写书评

暂无评论

SSRN

SSRN 2023年

作者： Zhu, Yongjun Kim, Donghun Jiang, Ting Zhao, Yi He, Jiangen Chen, Xinyi Lou, Wen Department of Library and Information Science Yonsei University Seoul Korea Republic of Department of Library and Information Science Sungkyunkwan University Seoul Korea Republic of School of Economics and Management Nanjing University of Science and Technology Nanjing China School of Information Sciences The University of Tennessee Knoxville United States Department of Cultural Media Yonsei University Seoul Korea Republic of Faculty of Economics and Management East China Normal University Shanghai China Key Laboratory of Advanced Theory and Application in Statistics and Data Science East China Normal University Ministry of Education of China Shanghai China

In this study, we examine the roles of dependency, reciprocity, and mentorship in the prediction of long-term research collaboration in five disciplines. We use co-authorship matrix-based multivariate time series features and interpretable machine learning to train long-term collaboration prediction models and interpret the feature importance of trained models. Overall, long-term research collaboration that is defined using various standards was rare across the examined disciplines, and the prediction results were moderate to good. We found dependency, reciprocity, and mentorship to have different roles in different disciplines. Among the three, mentorship was important in predicting long-term research collaboration in Agriculture, Geology, and Library and Information science. Reciprocity, which measures the interdependence between two researchers was important to prediction in the fields of Agriculture and Geology. Finally, dependency was important in all the disciplines with varying degrees of importance. © 2023, The Authors. All rights reserved.

关键词： Forecasting