检索结果-内蒙古大学图书馆

Spatial Distribution-Based Imbalanced Undersampling

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2023年第6期35卷 6376-6391页

作者： Yan, Yuanting Zhu, Yuanwei Liu, Ruiqing Zhang, Yiwen Zhang, Yanping Zhang, Ling Anhui Univ Sch Comp Sci & Technol Key Lab Intelligent Comp & Signal Proc Minist Educ Hefei 230601 Anhui Peoples R China

Undersampling is one of the most popular techniques for dealing with class-imbalance problems. Various undersampling methods have emerged over the past few decades. Each of them exhibits the superiority in some scenarios. However, selecting representative majority-class samples such that the structures of the selected groups are maintained according to the underlying imbalanced distribution remains a challenge. For this purpose, this paper proposes Spatial Distribution-based UnderSampling (SDUS) for imbalanced learning. SDUS uses a supervised constructive process to learn majority-class local patterns in terms of sphere neighborhoods (SPN). Two sample selection strategies, specifically, a top-down strategy and a bottom-up strategy, are proposed for maintaining the distribution pattern of original data in selecting majority-class sample subsets from different perspectives. SDUS introduces an ensemble technique that improves learning performance by utilizing the diversity caused by the randomness of the local-pattern learning process. Numerical experiments on 38 typical datasets from KEEL repository and 13 state-of-the-art comparison methods demonstrate the effectiveness of SDUS in maintaining the underlying distribution characteristics for imbalanced undersampling. The implementation of the proposed SDUS in programming language Python is available at https://***/ytyancp/SDUS.

关键词： Training Costs Biological neural networks Machine learning algorithms Clustering algorithms signal processing algorithms Proposals Undersampling imbalance local pattern sphere neighborhood ensemble learning

来源：评论

学校读者我要写书评

暂无评论

Regularized Optimal Transport Layers for Generalized Global Pooling Operations

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2023年第12期45卷 15426-15444页

作者： Xu, Hongteng Cheng, Minjie Renmin Univ Beijing Key Lab Big Data Management & Analy Meth Gaoling Sch Artificial Intelligence Beijing 100872 Peoples R China Renmin Univ Gaoling Sch Artificial Intelligence Beijing 100872 Peoples R China

Global pooling is one of the most significant operations in many machine learning models and tasks, which works for information fusion and structured data (like sets and graphs) representation. However, without solid mathematical fundamentals, its practical implementations often depend on empirical mechanisms and thus lead to sub-optimal, even unsatisfactory performance. In this work, we develop a novel and generalized global pooling framework through the lens of optimal transport. The proposed framework is interpretable from the perspective of expectation-maximization. Essentially, it aims at learning an optimal transport across sample indices and feature dimensions, making the corresponding pooling operation maximize the conditional expectation of input data. We demonstrate that most existing pooling methods are equivalent to solving a regularized optimal transport (ROT) problem with different specializations, and more sophisticated pooling operations can be implemented by hierarchically solving multiple ROT problems. Making the parameters of the ROT problem learnable, we develop a family of regularized optimal transport pooling (ROTP) layers. We implement the ROTP layers as a new kind of deep implicit layer. Their model architectures correspond to different optimization algorithms. We test our ROTP layers in several representative set-level machine learning scenarios, including multi-instance learning (MIL), graph classification, graph set representation, and image classification. Experimental results show that applying our ROTP layers can reduce the difficulty of the design and selection of global pooling - our ROTP layers may either imitate some existing global pooling methods or lead to some new pooling layers fitting data better.

关键词： Task analysis Optimization Neural networks Machine learning Feature extraction Solids signal processing algorithms Global pooling regularized optimal transport Bregman ADMM Sinkhorn scaling set representation graph embedding

来源：评论

学校读者我要写书评

暂无评论

An Efficient Parameter Optimization of Maximum Correntropy Criterion

引用

IEEE signal processing LETTERS 2023年 30卷 538-542页

作者： Shi, Long Shen, Lu Chen, Badong Sch Comp & Artificial Intelligence Chengdu 611130 Peoples R China Southwestern Univ Finance & Econ Financial Intelligence & Financial Engn Key Lab Si Chengdu 611130 Peoples R China Univ York Dept Elect Engn York YO10 5DD England Xi An Jiao Tong Univ Inst Artificial Intelligence & Robot Xian 710049 Peoples R China

The maximum correntropy criterion (MCC) algorithm depends upon two fundamental parameters, i.e., step-size and kernel width. Previous studies of parameter optimization in the MCC mainly focus on a single parameter (mainly the kernel width), lacking optimization research concerning both parameters. To this end, this letter investigates a novel optimization scheme simultaneously involving step-size and kernel width. The optimization framework is based on making the power of weight error vector undergo the steepest attenuation. Under the premise of maintaining the same evolutionary trend for time-varying step-size and kernel width, we formulate a constrained parameter optimization problem, where the step-size is subject to a kernel width induced constraint. By taking this approach, the original bivariate optimization can be transformed into a univariate optimization problem, which facilitates optimization solving. We further develop an existing reset scheme to make it suitable for kernel width to ensure a good tracking capability. In addition, we investigate the convergence behavior of the optimized algorithm. Simulation results demonstrate that the developed optimization scheme is beneficial for performance improvement, and the resulting algorithm outperforms some state-of-art MCC-based algorithms.

关键词： Kernel Optimization signal processing algorithms Market research Convergence Attenuation Heuristic algorithms Convergence behavior constrained parameter optimization kernel width maximum correntropy criterion tracking capability

来源：评论

学校读者我要写书评

暂无评论

Masked Spectral Bands Modeling With Shifted Windows: An Excellent Self-Supervised Learner for Classification of Medical Hyperspectral Images

引用

IEEE signal processing LETTERS 2023年 30卷 543-547页

作者： Li, Yuan Wu, Ruoyu Tan, Qijuan Yang, Zhengchun Huang, Hong Chongqing Univ Key Lab Optoelect Technol & Syst Educ Minist China Chongqing 400044 Peoples R China Chongqing Univ Dept Radiol Canc Hosp Chongqing 400030 Peoples R China Chongqing Canc Inst Chongqing 400030 Peoples R China Chongqing Canc Hosp Chongqing 400030 Peoples R China Chongqing Med Univ Women & Childrens Hosp Chongqing 401147 Peoples R China

Hyperspectral imaging has become a popular imaging technique in the medical field, and the development of algorithms for computer-aided diagnosis (CAD) is urgently required. Traditional deep learning techniques require a lot of annotated data, which is a burden on doctors. Self-supervised learning (SSL) is a solution for extracting feature representations from unlabeled data. However, traditional CNN-based SSL algorithms cannot explore relations between neighboring and long-range spectral bands, which limits classification performance. In this letter, the proposed solution is a novel SSL method using a transformer-based technique called masked spectral bands modeling with shifted windows (MSBMSW). This method predicts masked spectral bands as the pretext task and uses a self-attention mechanism with shifted windows to capture the divergence of neighboring spectral bands and enhance information exchange between long-range spectral bands. Experimental results demonstrate that MSBMSW achieves better classification results than many state-of-the-art methods and has potential clinical value for CAD of MHSIs.

关键词： Hyperspectral imaging Feature extraction Transformers Training Solid modeling signal processing algorithms Medical diagnostic imaging Medical hyperspectral images classification self-supervised learning transformer shift windows

来源：评论

学校读者我要写书评

暂无评论

Active-LATHE: An Active Learning Algorithm for Boosting the Error Exponent for Learning Homogeneous Ising Trees

引用

IEEE TRANSACTIONS ON INFORMATION THEORY 2023年第4期69卷 2537-2555页

作者： Zhang, Fengzhuo Tandon, Anshoo Tan, Vincent Y. F. Natl Univ Singapore Dept Elect & Comp Engn Singapore 119077 Singapore Indian Inst Sci IUDX Program Unit SID Bengaluru 560012 India Natl Univ Singapore Dept Math Singapore 119077 Singapore

The Chow-Liu algorithm (IEEE Trans. Inform. Theory, 1968) has been a mainstay for the learning of tree structured graphical models from i.i.d. sampled data vectors. Its theoretical properties have been well-studied and are well understood. In this paper, we focus on the class of trees that are arguably even more fundamental, namely homogeneous trees in which each pair of nodes that forms an edge has the same correlation ?. We ask whether we are able to further reduce the error probability of learning the structure of the homogeneous tree model when active learning is allowed. Our figure of merit is the error exponent, which quantifies the exponential rate of decay of the error probability with an increasing number of data samples. We design and analyze an algorithm ActiveLearning Algorithm for Trees with Homogeneous Edges (ACTIVE-LATHE), which surprisingly boosts the error exponent by at least 40% when ? is at least 0.8. For all other values of ?, we also observe commensurate, but more modest, improvements in the error exponent. Our analysis hinges on judiciously exploiting the minute but detectable statistical variation of the samples to allocate more data to parts of the graph in which we are less confident of being correct.

关键词： Graphical models Vegetation signal processing algorithms Correlation Error probability Random variables Task analysis Active learning graphical models ising models error exponents

来源：评论

学校读者我要写书评

暂无评论

Large Deviations for Products of Non-Identically Distributed Network Matrices With Applications to Communication-Efficient Distributed Learning and Inference

引用

IEEE TRANSACTIONS ON signal processing 2023年 71卷 1319-1333页

作者： Petrovic, Nemanja Bajovic, Dragana Kar, Soummya Jakovetic, Dusan Sahu, Anit Kumar Univ Novi Sad Fac Tech Sci Dept Fundamental Sci Novi Sad 21000 Serbia Univ Novi Sad Fac Tech Sci Dept Power Elect & Commun Engn Novi Sad 21000 Serbia Carnegie Mellon Univ Dept Elect & Comp Engn Pittsburgh PA 15213 USA Univ Novi Sad Fac Sci Dept Math & Informat Novi Sad 21000 Serbia Amazon Com Inc Alexa AI Pittsburgh PA 15232 USA

This paper studies products of independent but non-identically distributed random network matrices that arise as weight matrices in distributed consensus-type computation and inference procedures in peer-to-peer multi-agent networks. The non-identically distributed matrices studied in this paper model various application scenarios in which the agent communication network is time-varying, either naturally or engineered to achieve communication efficiency in computational procedures. First, under broad conditions on the statistics of the network matrix sequence, the product of the sequence is shown to converge almost surely to the consensus matrix and explicit large deviations rate of convergence are obtained. Specifically, given the admissible graph of interconnections modeling the base network topology, it is shown that the large deviations rate of consensus equals the minimum limiting value of the fluctuating graph cuts, where the edge costs are assigned through the current probabilities of the inter-agent communications. Secondly, an application of the above large deviations principle is studied in the context of distributed detection in time-varying networks with sequential observations. By adopting a consensus+innovations type distributed detection algorithm, as a by-product of this result, error exponents are obtained for the performance of distributed detection. It is shown that slow starts (slow increase) of inter-agent communication probabilities yield the same asymptotic error rate - and hence the same distributed detection performance, as if the communications were at their nominal levels from the beginning. As an important special case it is shown that when all the intermittent graph cuts have a link the probability of which increases to one, the performance of distributed detection is asymptotically optimal - i.e., equivalent to a centralized setup having access to all network data at all times.

关键词： Convergence Peer-to-peer computing Inference algorithms Electronic mail signal processing algorithms Limiting Estimation Distributed inference stochastic matrices error exponents inaccuracy rates large deviations consensus

来源：评论

学校读者我要写书评

暂无评论

Direct-Optimization-Based DC Dictionary Learning With the MCP Regularizer

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2023年第7期34卷 3568-3579页

作者： Li, Zhenni Yang, Zuyuan Zhao, Haoli Xie, Shengli Guangdong Univ Technol Sch Automat Guangzhou 510006 Peoples R China Guangdong HongKong Macao Joint Lab Smart Discrete Guangzhou 510006 Peoples R China Guangdong Key Lab IoT Informat Technol GDUT Guangzhou 510006 Peoples R China Sun Yat Sen Univ Sch Data & Comp Sci Guangzhou 510006 Peoples R China Minist Educ Key Lab Intelligent Informat Proc & Syst Integrat Guangzhou 510006 Peoples R China

Direct-optimization-based dictionary learning has attracted increasing attention for improving computational efficiency. However, the existing direct optimization scheme can only be applied to limited dictionary learning problems, and it remains an open problem to prove that the whole sequence obtained by the algorithm converges to a critical point of the objective function. In this article, we propose a novel direct-optimization-based dictionary learning algorithm using the minimax concave penalty (MCP) as a sparsity regularizer that can enforce strong sparsity and obtain accurate estimation. For solving the corresponding optimization problem, we first decompose the nonconvex MCP into two convex components. Then, we employ the difference of the convex functions algorithm and the nonconvex proximal-splitting algorithm to process the resulting subproblems. Thus, the direct optimization approach can be extended to a broader class of dictionary learning problems, even if the sparsity regularizer is nonconvex. In addition, the convergence guarantee for the proposed algorithm can be theoretically proven. Our numerical simulations demonstrate that the proposed algorithm has good convergence performances in different cases and robust dictionary-recovery capabilities. When applied to sparse approximations, the proposed approach can obtain sparser and less error estimation than the different sparsity regularizers in existing methods. In addition, the proposed algorithm has robustness in image denoising and key-frame extraction.

关键词： Machine learning Optimization Dictionaries Convergence Convex functions signal processing algorithms Approximation algorithms Convergence analysis dictionary learning direct optimization minimax concave penalty (MCP) regularizer

来源：评论

学校读者我要写书评

暂无评论

Accelerating Two-Phase Multiobjective Evolutionary Algorithm for Electric Location-Routing Problems

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2023年第9期24卷 9121-9136页

作者： Wang, Chao Liu, Rongrong Su, Yansen Zhang, Xingyi Anhui Univ Minist Educ Sch Artificial Intelligence Informat Mat & Intelligent Sensing Lab Anhui Prov Hefei Peoples R China Anhui Univ Engn Res Ctr Autonomous Unmanned Syst Technol Sch Artificial Intelligence Minist Educ Hefei 230601 Peoples R China Anhui Univ Minist Educ Key Lab Intelligent Comp & Signal Proc Hefei 230601 Peoples R China Anhui Univ Sch Artificial Intelligence Hefei 230601 Peoples R China

Electric location-routing problem is a challenging problem consisting of the optimization of electric vehicle routing and charging facility location, simultaneously. Existing algorithms generally adopt the two-phase search strategy to alternately optimize the routing and the location. However, they are usually criticized for the inefficiency as the problem scale increases. In order to improve the search efficiency in each phase, we propose an accelerating two-phase multiobjective evolutionary algorithm, where the learning method is used to mine the useful information from the historical search process to generate the high-quality routing and location offspring. To be specific, in the routing optimization phase, an interpolation method is developed to extract the frequent visiting orders existing in the historical best solutions. These frequent visiting orders are used to create potential routing offspring that can accelerate the convergence toward the optimal solutions. In the location optimization phase, a surrogate model is used to approximatively represent the relationship from routing to location, which can directly output a promising location scheme for a given routing offspring and thus reduce the optimization time. Experimental results on different scales of test instances demonstrate the competitiveness of the proposed algorithm in comparison with several state of-the-art algorithms, including four widely used heuristic algorithms and two multiobjective evolutionary algorithms.

关键词： Routing Search problems Optimization Heuristic algorithms Evolutionary computation Batteries signal processing algorithms Electric location-routing problem multiobjective optimization evolutionary algorithm interpolation surrogate

来源：评论

学校读者我要写书评

暂无评论

New Robust Sparse Convolutional Coding Inversion Algorithm for Ground Penetrating Radar Images

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2023年 61卷 1页

作者： Gallet, Matthieu Mian, Ammar Ginolhac, Guillaume Ollila, Esa Stelzenmuller, Nickolas Univ Savoie Mont Blanc LIST EA3703 F-74940 Annecy France Aalto Univ Dept Signal Proc & Acoust Aalto 00076 Finland Geolithe F-38920 Crolles France

In this article, we propose two algorithms to enhance the interpretability of the hyperbola in B-scans obtained with a ground penetrating radar (GPR). These hyperbolas are the responses of buried objects or cavities. To correctly detect and classify them, denoising is typically necessary for GPR images as the signal-to-noise ratio (SNR) is low, and the various interfaces naturally present in the Earth have a strong response. Both algorithms are based on a sparse convolutional coding model plus a low-rank component. It is solved through an alternating direction method of multipliers (ADMM) framework. In order to take into account the presence of outliers and the artifacts caused by the acquisition, the second algorithm is based on the Huber norm instead of the classic L-2 -norm. These algorithms are tested on a real dataset labeled by geophysicists. The results show the denoising efficiency of this approach, and in particular the robustness of the second algorithm.

关键词： Dictionaries Radar signal to noise ratio Radar imaging signal processing algorithms Shape Radar antennas Convolutive model ground penetrating radar (GPR) robust methods sparse inversion

来源：评论

学校读者我要写书评

暂无评论

DOA Estimation Based on Coherent Integration-Sparse Bayesian Learning With Time-Variant Gain-Phase Errors

引用

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS 2023年第6期59卷 7951-7962页

作者： Lu, Chaojie Zheng, Jibin Yang, Tianyuan Chang, Rui Liu, Hongwei Xidian Univ Natl Lab Radar Signal Proc Xian 710071 Peoples R China Chinese Acad Sci Natl Key Lab Microwave Imaging Technol Aerosp Informat Res Inst Beijing 100045 Peoples R China

Direction of arrival (DOA) estimation is an important topic of array signal processing and channels without gain-phase errors or with time-invariant gain-phase errors are assumed in most of the DOA estimation algorithms. However, with the measured data of a reconnaissance receiver system, we verify that the gain-phase error between channels is time-variation. Based on this fact, a more practical signal model with time-variant gain-phase errors is constructed. To obtain accurate DOA estimation, an algorithm termed as coherent integration-sparse Bayesian learning with time-variant gain-phase errors is (CI-SBLVGPE) devised which has two superiorities, i.e., 1) the CI operation integrates the signal energy into a single snapshot, which decreases the gain-phase errors, makes the compensation possible, and guarantees a low computational complexity;2) the priori information of the gain-phase errors, source DOA, and noise is used to reduce the computational complexity and improve the robustness. With mathematical analyses, numerical simulations, and real data processing, we validate the constructed model and superiorities of the proposed method.

关键词： Direction-of-arrival estimation Estimation signal processing algorithms Bayes methods Phased arrays Data models Measurement uncertainty Coherent integration DOA estimation gain-phase errors sparse Bayesian learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：