检索结果-内蒙古大学图书馆

International conference on image processing, Computer Vision and Machine Learning (ICICML)

作者： Chang, Haipeng He, Zhaokai Liu, Cunye Tao, Siyang Xu, Jaiden Li, Albert Chengzhe Zeng, Yijie Shanghai Amer Sch Puxi High Sch Div Shanghai Peoples R China Syracuse Univ Syracuse NY 13244 USA Suzhou Foreign Language Sch Suzhou Peoples R China United World Coll Changshu China Suzhou Peoples R China Alcanta Int Coll Nansha Guangzhou Guangzhou Peoples R China Cambridge Sch Weston Affiliat Weston FL USA

ISBN: (纸本)9781665464680

The predictions of text classifiers are often driven by spurious correlations - e.g., the term Spielberg may correlate with positive movie reviews, but the term itself does not convey a positive meaning. In this paper, we proposed alternative methods to distinguish spurious and genuine correlation in text classifiers based on previous work. We use logistic regression and MLP to replace matching in training word classifier and get the ITE for each method. Also, we aim to show the role of ITE in causal inference tasks. Experiments on two datasets (IMDB reviews and Kindle reviews) suggest that logistic regression and MLP perform better for feature selection on larger datasets, and the use of ITE can improve the accuracy for identifying spurious correlation.

关键词： Calculation ITE Correlation

来源：评论

学校读者我要写书评

暂无评论

SelfDeblur with Sparsity Enforced Bregman Learning

SelfDeblur with Sparsity Enforced Bregman Learning

引用

International conference on Acoustics, Speech, and Signal processing (ICASSP)

作者： Yaoyun Zeng Beier Chen Hongxia Wang Department of Mathematics National University of Defense Technology Changsha China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Learning-driven methods have revolutionized the field of blind deconvolution, with SelfDeblur as a pioneering method. It uses deep image priors to jointly estimate blur kernels and latent clear images, employing deep neural networks to automatically learn these priors and effectively capture their statistical properties. Building on the success of SelfDeblur, this paper proposes a novel sparse optimization method, named B-SelfDeblur, which employs linearized Bregman iterations to train neural networks. Our approach addresses the challenges of processing large-scale blurry images by starting with minimal parameters and incrementally adding necessary ones in an inverse scale space manner. Additionally, we incorporate a smoothing technique for the total variation (TV) regularization term, which ensures our loss function is continuously differentiable. Experimental results demonstrate that B-SelfDeblur can significantly optimize network weight storage while maintaining deblurring effectiveness. Furthermore, we provide a partial theoretical analysis and ablation studies, which confirm that our loss function is monotonically decreasing and that our method is well-suited for alternating iterative optimization.

关键词： TV Deconvolution Smoothing methods Signal processing algorithms Optimization methods Signal processing Iterative methods Speech processing Kernel Convergence

来源：评论

学校读者我要写书评

暂无评论

Two Efficient Recursive Total Least Squares Solutions Based on the Grouping Strategy

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2024年 73卷

作者： Qi, Zhijun Wang, Wei Zeng, Wenxian Wei, Yelong Luo, Tao Fang, Xing Inst Adv Sci Facil Shenzhen 518107 Peoples R China Wuhan Univ Sch Geodesy & Geomat Wuhan 430079 Peoples R China Univ Sci & Technol China Natl Synchrotron Radiat Lab Hefei 230023 Peoples R China

Total least squares (TLS) has been comprehensively explored over the last few decades due to its excellent statistical properties. However, the unsatisfactory efficiency of TLS from the intricate process of inverting large matrices limits its practical applications, especially when processing big data. To address this issue, we propose two recursive TLS (RTLS) solutions based on the grouping strategy, where all the observations are partitioned into given independent subsets. Within approximated RTLS, observations in the new subset are recursively integrated with results from preceding subsets. The sequential solution is further extended to be exact RTLS. Both approximated RTLS and exact RTLS can significantly reduce the matrix dimensions in the calculation. The proposed methods are compared with several existing algorithms with the experiments of 2-D line fitting, image registration, and reference network transformation. In the three experiments, our solutions improve the efficiency of TLS by more than one order of magnitude.

关键词： Vectors Symmetric matrices Mathematical models Fitting Linear matrix inequalities Computational modeling stochastic processes Complexity theory Big Data Accuracy Algorithm efficiency coordinate transformation errors-in-variables (EIVs) model grouping strategy image registration line fitting recursive total least squares (RTLS)

来源：评论

学校读者我要写书评

暂无评论

Enhancing Multi-view Mammography image Classification: By using Breast Region Extraction Method and statistical Features

Enhancing Multi-view Mammography Image Classification: By us...

引用

Iranian Machine Vision and image processing (MVIP)

作者： Neda Shirany Bidabadi Elham Mahmoudzadeh Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan Iran

Breast cancer is one of the most dangerous diseases among women. Different methods are used to diagnose this cancer that among these, imaging and computer-aided systems are more common. In these systems, one of the most important step is preprocessing and removing unnecessary areas of the images, as well as extracting the chest area. In this paper, we present a method that consists of preprocessing, feature extraction, and using a machine learning classifier. In the preprocessing step, we propose a method to extract the region of interest in both angles of mammography images. The proposed novel method includes applying gamma correction thresholding to the images and obtaining two binary images based on the proposed threshold using the Otsu method. Results show the proposed method successfully removes the chest muscle with 98% accuracy. In the next, for feature extraction phase, we utilize three different methods for extracting features. Finally, by employing an Extra tree model classifier, we classify mammography images into normal and abnormal. By incorporating the block-based feature extraction method, we achieve 98% accuracy in classification. Overall, our approach demonstrates the effectiveness of preprocessing and feature extraction for diagnosing breast cancer using mammography images.

关键词：

来源：评论

学校读者我要写书评

暂无评论

stochastic Majorize-Minimize Subspace Algorithm with Application to Binary Classification 29

Stochastic Majorize-Minimize Subspace Algorithm with Applica...

引用

29th European Signal processing conference, EUSIPCO 2021

作者： Fest, Jean-Baptiste Chouzenoux, Émilie Université Paris-Saclay Inria CentraleSupélec Centre de Vision Numérique France

ISBN: (纸本)9789082797060

In a learning context, data distribution are usually unknown. Observation models are also sometimes complex. In an inverse problem setup, these facts often lead to the minimization of a loss function with uncertain analytic expression. Consequently, its gradient cannot be evaluated in an exact manner. These issues have has promoted the development of so-called stochastic optimization methods, which are able to cope with stochastic errors in the gradient term. A natural strategy is to start from a deterministic optimization approach as a baseline, and to incorporate a stabilization procedure (e.g., decreasing stepsize, averaging) that yields improved robustness to stochastic errors. In the context of large-scale, differentiable optimization, an important class of methods relies on the principle of majorization-minimization (MM). MM algorithms are becoming increasingly popular in signal/image processing [18], [36] and machine learning [27], [34], [38]. MM approaches are fast, stable, require limited manual settings, and are often preferred by practitioners in application domains such as medical imaging [16] and telecommunications [29]. The present work introduces novel theoretical convergence guarantees for MM algorithms when approximate gradient terms are employed, generalizing some recent work [11], [27] to a wider class of functions and algorithms. We illustrate our theoretical results with a binary classification problem. © 2021 European Signal processing conference. All rights reserved.

关键词： Optimization

来源：评论

学校读者我要写书评

暂无评论

The Study of methods to Improve the Accuracy of Glycemic Control Based on statistical methods 5

The Study of Methods to Improve the Accuracy of Glycemic Con...

引用

5th International conference on Vision, image and Signal processing (ICVISP)

作者： Terenteva, Arina Limorenko, Maria Bauman Moscow State Tech Univ Dept Robot & Complex Automat Moscow Russia Russian Acad Sci Dept Theory Mech & MachinesMech Engn Res Inst Moscow Russia

ISBN: (纸本)9781665407700

Erroneous determination of blood glucose concentration in patients with diabetes can lead to errors in treatment. To prescribe the correct and non-traumatic treatment to the patient, you need to increase the accuracy of determining the concentration of glucose in the patient's blood. The authors of the article suggest the use of adaptive statistical methods to improve accuracy, which work well with a small number of repeated measurements under conditions of uncertainty of external factors. Such methods have proven themselves well in various fields of science and technology, where precise control of a large group of measurements is required. Such statistical methods have been well studied in terms of mathematics and applied in high-precision manufacturing. They allow you to adjust the required dose of the injected substance on the basis of several previous measurements in order to exclude insufficient or excessive administration. Both of these cases - an insufficient and excessive amount of the drug - lead to an incorrect glucose content in the patient's blood (hypo- and hyperglycemia), and have a bad effect on further treatment. To exclude such cases and improve the quality of treatment and the future life of patients with diabetes mellitus, a solution method using statistics is proposed.

关键词： glucose control accuracy diabetes moving average

来源：评论

学校读者我要写书评

暂无评论

Skeletal Modeling of the Tricuspid Valve via Cylindrical Parameterization

Skeletal Modeling of the Tricuspid Valve via Cylindrical Par...

引用

conference on Medical Imaging - image processing

作者： Vicory, Jared Herz, Christian Flynn, Maura Cianciulli, Alana Sabin, Patricia Lasso, Andras Jolley, Matthew A. Paniagua, Beatriz Kitware Inc Carrboro NC 27510 USA Childrens Hosp Philadelphia Dept Anesthesia & Crit Care Med Philadelphia PA USA Childrens Hosp Philadelphia Div Pediat Cardiol Philadelphia PA USA Queens Univ Kingston ON Canada

ISBN: (纸本)9781510660335;9781510660342

Hypoplastic left heart syndrome is a severe congenital heart defect requiring surgical intervention shortly after birth. The surgery is complex and often leads to complications requiring additional surgeries. Understanding the relationship between the structure of the tricuspid valve and functional complications could lead to more powerful diagnostic and treatment options. Because the tricuspid valve does not have spherical topology, many traditional methods for creating boundary-based or skeleton-based shape models that require spherical parameterization of an object are not applicable unless individual leaflets are independently parameterized and then merged in a multi-object model. Instead we propose to create skeletal models (s-reps) of the entire tricuspid valve structure using a cylindrical parameterization. We modify a traditional cylindrical parameterization approach by adaptively changing angle sampling based on landmarks to produce anatomically relevant correspondence across a population of objects. From this we derive s-reps which yield an improved shape space and classification performance compared with previous approaches.

关键词： statistical shape modeling skeletal models cylindrical parameterization tricuspid valve

来源：评论

学校读者我要写书评

暂无评论

statistical characterization of segregation-driven inhomogeneities in metallic microstructures employing fast first-order variograms

引用

MATERIALS TODAY COMMUNICATIONS 2023年 34卷

作者： Benito, Santiago Egels, Gero Hartmaier, Alexander Weber, Sebastian Ruhr Univ Bochum Chair Mat Technol D-44780 Bochum Germany Ruhr Univ Bochum Chair Micromech & Macroscop Modelling ICAMS D-44780 Bochum Germany

The microstructure is the centerpiece connecting thermodynamic, compositional, and kinetic stochasticity with macroscopic behavior. As such, its thorough description is of fundamental importance: Microscopical spatial composition fluctuations can critically undermine or improve material performance. Still, both traditional and modern, state-of-the-art statistical microstructural characterization methods overlook micro and mesosegre-gations. Instead, they generally focus on microconstituent and grain examination. Segregation effects are thus commonly described on a case-by-case basis or employing parameters that lack spatial interpretation. We propose fast first-order variograms as a convenient statistical tool to comprehensively describe chemical segregations in metallic materials. First-order variograms are physically meaningful descriptors capable of revealing spatial variations and correlations. In particular, we discuss the derivation, application, advantages, and limits of their fast computation using the fast Fourier transform, which brings a substantial speed increase over the method-of-moments estimation. Furthermore, we compare them to popular texture characterization techniques borrowed from image processing and analysis. With this work, we establish a simple-to-use, yet powerful method to characterize the severeness of micro and mesosegregations and, thus, to quantify their influence on material behavior.

关键词： Microstructural characterization Geostatistics statistical description stochastic modeling Random fields Spatial data

来源：评论

学校读者我要写书评

暂无评论

Proceedings of the 25th International conference on Digital Audio Effects, DAFx20in22

Proceedings of the 25th International Conference on Digital ...

引用

25th International conference on Digital Audio Effects, DAFx 2022

ISBN: (纸本)9783200085992

The proceedings contain 42 papers. The topics discussed include: room impulse response estimation using signed distance functions;an acoustic paintbrush method for simulated spatial room impulse responses;flexible real-time reverberation synthesis with accurate parameter control;evaluation of a stochastic reverberation model based on the source image principle;identification of nonlinear circuits as port-Hamiltonian systems;an equivalent circuit interpretation of antiderivative antialiasing;non-iterative schemes for the simulation of nonlinear audio circuits;applications of port Hamiltonian methods to non-iterative stable simulations of the KORG35 and MOOG 4-pole VCF;deforming the oscillator: iterative phases over parametrizable closed paths;continuous state modeling for statistical spectral synthesis;subjective evaluation of sound quality and control of drum synthesis with Stylewavegan;and a low-latency quasi-linear-phase octave graphic equalizer.

关键词：

来源：评论

学校读者我要写书评

暂无评论

FIBER-SAMPLED stochastic MIRROR DESCENT FOR TENSOR DECOMPOSITION WITH β-DIVERGENCE

FIBER-SAMPLED STOCHASTIC MIRROR DESCENT FOR TENSOR DECOMPOSI...

引用

IEEE International conference on Acoustics, Speech and Signal processing (ICASSP)

作者： Pu, Wenqiang Ibrahim, Shahana Fu, Xiao Hong, Mingyi Chinese Univ Hong Kong Shenzhen Res Inst Big Data Shenzhen Peoples R China Oregon State Univ Sch Elect Engn & Comp Sci Corvallis OR 97331 USA Univ Minnesota Dept Elect & Comp Engn Minneapolis MN 55455 USA

ISBN: (纸本)9781728176055

Canonical polyadic decomposition (CPD) has been a workhorse for multimodal data analytics. This work puts forth a stochastic algorithmic framework for CPD under beta-divergence, which is well-motivated in statistical learning-where the Euclidean distance is typically not preferred. Despite the existence of a series of prior works addressing this topic, pressing computational and theoretical challenges, e.g., scalability and convergence issues, still remain. In this paper, a unified stochastic mirror descent framework is developed for large-scale beta-divergence CPD. Our key contribution is the integrated design of a tensor fiber sampling strategy and a flexible stochastic Bregman divergence-based mirror descent iterative procedure, which significantly reduces the computation and memory cost per iteration for various beta. Leveraging the fiber sampling scheme and the multilinear algebraic structure of low-rank tensors, the proposed lightweight algorithm also ensures global convergence to a stationary point under mild conditions. Numerical results on synthetic and real data show that our framework attains significant computational saving compared with state-of-the-art methods.

关键词： Tensor decomposition beta-divergence stochastic optimization mirror descent method

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：