检索结果-内蒙古大学图书馆

arXiv 2019年

作者： Li, Bo Xu, Kele Feng, Dawei Mi, Haibo Wang, Huaimin Zhu, Jian Beijing University of Posts and Telecommunications Automation Dept. Beijing100876 China National Key Lab of Parallel and Distributed Processing Changsha China National University of Defense Technology Changsha China Department of Linguistics University of Michigan Ann Arbor United States

B-mode ultrasound tongue imaging is widely used in the speech production field. However, efficient interpretation is in a great need for the tongue image sequences. Inspired by the recent success of unsupervised deep learning approach, we explore unsupervised convolutional network architecture for the feature extraction in the ultrasound tongue image, which can be helpful for the clinical linguist and phonetics. By quantitative comparison between different unsupervised feature extraction approaches, the denoising convolutional autoencoder (DCAE)-based method outperforms the other feature extraction methods on the reconstruction task and the 2010 silent speech interface challenge. A Word Error Rate of 6.17% is obtained with DCAE, compared to the state-of-the-art value of 6.45% using Discrete cosine transform as the feature extractor. Our codes are available at https://***/DeePBluE666/Source-code1. Copyright © 2019, The Authors. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Efficient Training of Convolutional Neural Nets on Large distributed Systems

Efficient Training of Convolutional Neural Nets on Large Dis...

引用

IEEE International Conference on Cluster Computing (CLUSTER)

作者： Sreedhar, Dheeraj Saxena, Vaibhav Sabharwal, Yogish Verma, Ashish Kumar, Sameer IBM Res Bangalore Karnataka India Google Inc Sunnyvale CA 94089 USA

ISBN: (纸本)9781538683194

Deep Neural Networks (DNNs) have achieved impressive accuracy in many application domains including image classification. Training of DNNs is an extremely compute-intensive process and is solved using variants of the stochastic gradient descent (SGD) algorithm. A lot of recent research has focused on improving the performance of DNN training. In this paper, we present optimization techniques to improve the performance of the data parallel synchronous SGD algorithm using the Torch framework: (i) we maintain data in-memory to avoid file I/O overheads, (ii) we propose optimizations to the Torch data parallel table framework that handles multithreading, and (iii) we present MPI optimization to minimize communication overheads. We evaluate the performance of our optimizations on a Power 8 Minsky cluster with 64 nodes and 256 NVidia Pascal P100 GPUs. With our optimizations, we are able to train 90 epochs of the ResNet-50 model on the imagenet-1k dataset using 256 GPUs in just 48 minutes. This significantly improves on the previously best known performance of training 90 epochs of the ResNet-50 model on the same dataset using the same number of GPUs in 65 minutes. To the best of our knowledge, this is the best known training performance demonstrated for the imagenet-1k dataset using 256 GPUs.

关键词： Training Optimization Computational modeling Graphics processing units distributed databases Neural networks Peer-to-peer computing

来源：评论

学校读者我要写书评

暂无评论

Underwater image Enhancement with a Total Generalized Variation Illumination Prior

Underwater Image Enhancement with a Total Generalized Variat...

引用

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Zhengjie Zhao Yuxiang Dai Peixian Zhuang School of Electronic and Information Engineering Nanjing University of Information Science and Technology Nanjing China

ISBN: (数字)9781728143286

ISBN: (纸本)9781728143293

In order to promote the visual quality of degraded underwater images, we develop an innovative variational model based on Retinex with a total generalized variation (TGV) prior on the illumination. The TGV prior is adopted to approximate piecewise smoothness and piecewise linear smoothness of the illumination, which combines the first-order and second-order total variation (TV) to model the variation of illumination. When adopting this illumination prior in the Retinex-based algorithm, the illumination and reflection are well separated, and underwater enhanced results appear more natural and their details and edges are better preserved. Then an efficient iterative optimization method is derived to settle the proposed model via alternately calculating the illumination and the reflection simultaneously. Numerous experiments on both visual results and objective metrics demonstrate the superiority of our method compared with several underwater enhancement methods. In addition, the proposed method can be extended for dehazing, sandstorm removal and low illumination image enhancement, which can illustrate better capacity of our model.

关键词： Lighting image color analysis Mathematical model Optimization image enhancement Visualization Adaptation models

来源：评论

学校读者我要写书评

暂无评论

Underwater image Enhancement Based on Dehazing and Color Correction

Underwater Image Enhancement Based on Dehazing and Color Cor...

引用

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Hanyu Li Peixian Zhuang Wang Wei Jingjing Li Nanjing University of Information Science and Technology Nanjing China

ISBN: (数字)9781728143286

ISBN: (纸本)9781728143293

Clear and clean underwater images can provide valuable information, which are crucial for developing, exploring, and protecting the underwater resources. However, the raw underwater image seldom fulfills the requirements concerning the underwater research due to the serious image degradation. To enhance single underwater image, we propose an effective and new method combined on the dehazing and color correction algorithm. First, we obtain the dehazed image by a fusion dehazing method calculated on the difference between maximum green-blue dark channels and maximum red dark channel, and the top 0:1% bright pixels in green-blue dark channels. Second, we obtain the enhanced image via the color restoration method corresponding to the human visual system. Finally, we use an efficient and simple weight fusion strategy to incorporate the dehazed image and the enhanced image for yielding the final high quality underwater image. Subjective assessment demonstrates that our method can remove haze, correct color cast, improve image brightness, and preserve image naturalness. The enhanced results obtain the highest values in terms of entropy, average gradient, UCIQE, UIQM, and CCF, which is superior to several existing underwater enhancement methods. Moreover, our method can improve other degraded image quality, such as low-light image and haze image, and can benefit the application tests, such as key points matching and edge detection.

关键词： image color analysis Degradation image restoration image edge detection image quality Channel estimation image enhancement

来源：评论

学校读者我要写书评

暂无评论

FVR-SGD: A New Flexible Variance-Reduction Method for SGD on Large-Scale Datasets 25th

FVR-SGD: A New Flexible Variance-Reduction Method for SGD on...

引用

25th International Conference on Neural Information processing (ICONIP)

作者： Tang, Mingxing Huang, Zhen Qiao, Linbo Du, Shuyang Peng, Yuxing Wang, Changjian Natl Univ Def Technol Sci & Technol Parallel & Distributed Lab Changsha 410073 Peoples R China Tongji Univ Shanghai Peoples R China

ISBN: (纸本)9783030041793;9783030041786

Stochastic gradient descent (SGD) is a popular optimization method widely-used in machine learning, while the variance of gradient estimation leads to slow convergence. To accelerate the speed, many variance reduction methods have been proposed. However, most of these methods require additional memory cost or computational burden on full gradient, which results in low efficiency or even unavailable while applied to real-world applications with large-scale datasets. To handle this issue, we propose a new flexible variance reduction method for SGD, named FVR-SGD, which can reduce memory overhead and speedup the convergence using flexible subset size without extra operation. The details of convergence property are presented, the convergence of variance reduction method using flexible subset size can be guaranteed. Several numerical experiments are conducted on a genre of real-world large-scale datasets. The experimental results demonstrated that FVR-SGD outperforms contemporary SVRG algorithm. Specifically, the proposed method can achieve up to 40% reduction in the training time to solve the optimization problem of logistic regression.

关键词： Machine learning Optimization Stochastic gradient descent Variance reduction distributed optimization

来源：评论

学校读者我要写书评

暂无评论

Group Importance Sampling for particle filtering and MCMC

引用

DIGITAL SIGNAL processing 2018年 82卷 133-151页

作者： Martino, Luca Elvira, Victor Camps-Valls, Gustau Univ Valencia Image Proc Lab Valencia Spain Univ Carlos III Madrid Dept Stat Signal Proc Madrid Spain IMT Lille Douai CRISTAL UMR 9189 Villeneuve Dascq France

Bayesian methods and their implementations by means of sophisticated Monte Carlo techniques have become very popular in signal processing over the last years. Importance Sampling (IS) is a wellknown Monte Carlo technique that approximates integrals involving a posterior distribution by means of weighted samples. In this work, we study the assignation of a single weighted sample which compresses the information contained in a population of weighted samples. Part of the theory that we present as Group Importance Sampling (GIS) has been employed implicitly in different works in the literature. The provided analysis yields several theoretical and practical consequences. For instance, we discuss the application of GIS into the Sequential Importance Resampling framework and show that Independent Multiple Try Metropolis schemes can be interpreted as a standard Metropolis-Hastings algorithm, following the GIS approach. We also introduce two novel Markov Chain Monte Carlo (MCMC) techniques based on GIS. The first one, named Group Metropolis Sampling method, produces a Markov chain of sets of weighted samples. All these sets are then employed for obtaining a unique global estimator. The second one is the distributed Particle Metropolis-Hastings technique, where different parallel particle filters are jointly used to drive an MCMC algorithm. Different resampled trajectories are compared and then tested with a proper acceptance probability. The novel schemes are tested in different numerical experiments such as learning the hyperparameters of Gaussian Processes, two localization problems in a wireless sensor network (with synthetic and real data) and the tracking of vegetation parameters given satellite observations, where they are compared with several benchmark Monte Carlo techniques. Three illustrative Matlab demos are also provided. (C) 2018 Elsevier Inc. All rights reserved.

关键词： Importance Sampling Markov Chain Monte Carlo (MCMC) Particle filtering Particle Metropolis-Hastings Multiple Try Metropolis Bayesian inference

来源：评论

学校读者我要写书评

暂无评论

A comparison study on nonlinear dimension reduction methods with kernel variations: Visualization, optimization and classification

arXiv

引用

arXiv 2019年

作者： Kempfert, Katherine C. Wang, Yishi Chen, Cuixian Wong, Samuel W.K. University of Florida GainesvilleFL32611 United States University of North Carolina Wilmington University of Waterloo

Because of high dimensionality, correlation among covariates, and noise contained in data, dimension reduction (DR) techniques are often employed to the application of machine learning algorithms. Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and their kernel variants (KPCA, KLDA) are among the most popular DR methods. Recently, Supervised Kernel Principal Component Analysis (SKPCA) has been shown as another successful alternative. In this paper, brief reviews of these popular techniques are presented first. We then conduct a comparative performance study based on three simulated datasets, after which the performance of the techniques are evaluated through application to a pattern recognition problem in face image analysis. The gender classification problem is considered on MORPH-ii and FG-NET, two popular longitudinal face aging databases. Several feature extraction methods are used, including biologically-inspired features (BIF), local binary patterns (LBP), histogram of oriented gradients (HOG), and the Active Appearance Model (AAM). After applications of DR methods, a linear support vector machine (SVM) is deployed with gender classification accuracy rates exceeding 95% on MORPH-ii, competitive with benchmark results. A parallel computational approach is also proposed, attaining faster processing speeds and similar recognition rates on MORPH-ii. Our computational approach can be applied to practical gender classification systems and generalized to other face analysis tasks, such as race classification and age prediction. Copyright © 2019, The Authors. All rights reserved.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

distributed eigenfaces for massive face image data

引用

MULTIMEDIA TOOLS AND APPLICATIONS 2017年第24期76卷 25983-26000页

作者： Park, Jeong-Keun Park, Ho-Hyun Park, Jaehwa Chung Ang Univ Dept Comp Sci & Engn Seoul South Korea Chung Ang Univ Dept Elect & Elect Engn Seoul South Korea

The assumption that the number of training samples is less than the number of pixels in a face image is essential for conventional eigenface-based face recognition. But recently, it has become impractical for massive face image collections. A parallel processing method using distributed eigenfaces is presented. A massive face image set was divided into a bunch of small subsets that satisfied the assumption of conventional approaches. Eigenfaces were extracted from the subsets and stored in a cloud system. Face recognition was performed by parallel processing using the distributed eigenfaces in the cloud system. A face recognition system was implemented in the Hadoop system. Various experiments were performed to test the validity of the distributed eigenface-based approach. The experimental results show that, compared to conventional methods, the implemented distributed face recognition system worked well for large datasets without significant performance degradation.

关键词： Eigenface Face recognition parallel processing Hadoop

来源：评论

学校读者我要写书评

暂无评论

image Matching Algorithm Based on GMS

Image Matching Algorithm Based on GMS

引用

IEEE International Symposium on parallel and distributed processing with Applications and IEEE International Conference on Ubiquitous Computing and Communications (ISPA/IUCC)

作者： Hua Li Wen Liu Cheng Han Xiangxu Kong School of Computer Science and Technology Changchun University of Science and Technology Changchun China Changchun University of Science and Technology Changchun China

ISBN: (纸本)9781728152103

image matching is a technique for measuring the similarity of two or more images to determine the correspondence between images. It is the main research content in the field of computer vision. In order to solve the problem of low matching accuracy of Grid-based Motion Statistics(GMS) algorithm in complex situations such as illumination changes, rotation changes or scale changes, an improved GMS image matching algorithm is proposed. Firstly, we use AKAZE to detect and describe the feature points. Then, the bi-directional KNN matching algorithm is adopted for rough matching. Finally, the GMS algorithm is used to filter and eliminate the false matching point pairs. In order to verify the effectiveness of the proposed method, the Oxford standard datasets and practical images are experimented. The experimental results show that the correct matching rate of the proposed algorithm can reach to 90%. Our method can match better in complex scenes than ORB, and GMS algorithms. In terms of matching time, the proposed method reduces about 15% than original GMS.

关键词： computer vision feature extraction image matching image motion analysis nearest neighbour methods image matching images computer vision image motion analysis Feature extraction nearest neighbor algorithm algorithms Object Request Broker PPP1R3A gene

来源：评论

学校读者我要写书评

暂无评论

Meta-Learning Techniques to Analyze the Raman Data for Optical Diagnosis of Oral Cancer Detection

Meta-Learning Techniques to Analyze the Raman Data for Optic...

引用

IEEE International Symposium on parallel and distributed processing with Applications and IEEE International Conference on Ubiquitous Computing and Communications (ISPA/IUCC)

作者： Mukta Sharma Lokesh Sharma Ming-Jer Jeng Liann-Be Chang Shiang-Fu Huang Shih-Lin Wu Department of Electronic Engineering Chang Gung University Taoyuan Taiwan AI Innovation Research Center Chang Gung University Taoyuan Department of Otolaryngology-Head and Neck Surgery Chang Gung Memorial Hospital Linkou Taiwan Department of Public Health Chang Gung University Taoyuan Taiwan Department of Cardiology Chang Gung Memorial Hospital Taoyuan Department of Electrical Engineering Ming Chi University of technology New Taipei City Taiwan Department of Computer Science and Information Engineering Chang Gung University Taoyuan Taiwan

ISBN: (纸本)9781728152103

Recently, Machine Learning methods have shown great improvement while analyzing the biomedical data. Raman Spectroscopy (RS), a non-invasive technique, and widely used in screening to diagnose the oral cancer. In order to spot cancer in a smarter and faster way, we have employed Meta-Learning (ML) techniques to learn such as Bagging and Boosting on RS data. Further, we employed normal and tumor tissue class classification by Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and Adaptive Boosting (AdaBoost) classifiers. The present study aims at examining the RS data with total 110 samples, including 57 tumor and 53 normal ones. To evaluate the performance, we have used the training samples to optimize, and testing samples to generalize the model parameters. The results show that the AdaBoost classifier with Bagging techniques showed the significant changes in accuracy.

关键词： biomedical optical imaging cancer image classification learning (artificial intelligence) medical image processing tumours Biomedical optical imaging Medical image processing image classification Tumours Mouth Neoplasms Malignant Oral Cavity Tumor cancer test OPTICAL DIAGNOSTICS Boost bagging PURSE

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：