检索结果-内蒙古大学图书馆

An Effective Method for Modeling Two-dimensional Sky Background of LAMOST

学校读者我要写书评

暂无评论

Proceedings of the International Astronomical Union 2017年第S325期12卷 63-66页

作者： Hasitieer Haerken Fuqing Duan Jiannan Zhang Ping Guo Image Processing and Pattern Recognition Laboratory Beijing Normal University 100875 Beijing China email: fqduan@bnu.edu.cnpguo@*** National Astronomical Observatories & Chinese Academy of Sciences 100012 Beijing China email: hastear@***jnzhang@***

Each CCD of LAMOST accommodates 250 spectra, while about 40 are used to observe sky background during real observations. How to estimate the unknown sky background information hidden in the observed 210 celestial spectra by using the known 40 sky spectra is the problem we solve. In order to model the sky background, usually a pre-observation is performed with all fibers observing sky background. We use the observed 250 skylight spectra as training data, where those observed by the 40 fibers are considered as a base vector set. The Locality-constrained Linear Coding (LLC) technique is utilized to represent the skylight spectra observed by the 210 fibers with the base vector set. We also segment each spectrum into small parts, and establish the local sky background model for each part. Experimental results validate the proposed method, and show the local model is better than the global model.

关键词： LAMOST celestial spectra sky background modeling

On Measuring and Quantifying Performance: Error Rates, Surrogate Loss, and an Example in SSL

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Loog, Marco Krijthe, Jesse Jensen, Are C. Department of Molecular Epidemiology Leiden University Medical Center Netherlands Pattern Recognition Laboratory Delft University of Technology Netherlands Image Section University of Copenhagen Denmark Department of Informatics University of Oslo Norway

In various approaches to learning, notably in domain adaptation, active learning, learning under covariate shift, semi-supervised learning, learning with concept drift, and the like, one often wants to compare a baseline classifier to one or more advanced (or at least different) strategies. In this chapter, we basically argue that if such classifiers, in their respective training phases, optimize a so-called surrogate loss that it may also be valuable to compare the behavior of this loss on the test set, next to the regular classification error rate. It can provide us with an additional view on the classifiers’ relative performances that error rates cannot capture. As an example, limited but convincing empirical results demonstrates that we may be able to find semi-supervised learning strategies that can guarantee performance improvements with increasing numbers of unlabeled data in terms of log-likelihood. In contrast, the latter may be impossible to guarantee for the classification error rate. Copyright © 2017, The Authors. All rights reserved.

关键词： Supervised learning

Fast KNN search for big data with set compression tree and best bin first 2

学校读者我要写书评

暂无评论

Fast KNN search for big data with set compression tree and b...

2nd International Conference on Cloud Computing and Internet of Things, CCIOT 2016

作者： Chen, Zhenjie Yan, Jingqi Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University Key Laboratory of Ministry of Education for System Control and Information Processing China Shanghai China

ISBN: (纸本)9781467398213

This paper proposes k nearest neighbors (kNN) search based on set compression tree (SCT) and best bin first (BBF) to deal with the problem for big data. The large compression rate by set compression tree is achieved by compressing the set of descriptors jointly instead of compressing on a per-descript or basis. So set compression tree has a good performance in kNN search at a low bit rate. At the same time, the best bin first (BBF) is a very efficient algorithm to find the approximately kNN from a large number of high dimensional feature descriptors. SCT-BBF is a novel exploration and it improves search performance in three aspects: First, SCT-BBF requires less memory footprint, which is important in big data age. Second, it increases accuracy compared traditional method like KD-Tree and original SCT. SCT-BBF can be used with other data processing methods like PCA and SIFT to perform better. Third, this paper adopts finer search to increase accuracy at a slight loss of speed. And it can extend to big data easily. © 2016 IEEE.

关键词： Big data

Varifocal-Net: A Chromosome Classification Approach Using Deep Convolutional Networks

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Qin, Yulei Wen, Juan Zheng, Hao Huang, Xiaolin Yang, Jie Song, Ning Zhu, Yue-Min Wu, Lingqian Yang, Guang-Zhong Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University Shanghai200240 China Center for Medical Genetics School of Life Sciences Central South University Changsha410078 China Shanghai Key Laboratory of Reproductive Medicine School of Medicine Shanghai Jiao Tong University Shanghai200025 China Diagens-Hangzhou Hangzhou311121 China University Lyon INSA Lyon CNRS INSERM CREATIS UMR 5220 U1206F-69621 France Hamlyn Centre for Robotic Surgery Imperial College London SW72AZ United Kingdom

Chromosome classification is critical for karyotyping in abnormality diagnosis. To expedite the diagnosis, we present a novel method named Varifocal-Net for simultaneous classification of chromosomes type and polarity using deep convolutional networks. The approach consists of one global-scale network (G-Net) and one local-scale network (L-Net). It follows three stages. The first stage is to learn both global and local features. We extract global features and detect finer local regions via the G-Net. By proposing a varifocal mechanism, we zoom into local parts and extract local features via the L-Net. Residual learning and multi-task learning strategies are utilized to promote high-level feature extraction. The detection of discriminative local parts is fulfilled by a localization subnet of the G-Net, whose training process involves both supervised and weakly-supervised learning. The second stage is to build two multi-layer perceptron classifiers that exploit features of both two scales to boost classification performance. The third stage is to introduce a dispatch strategy of assigning each chromosome to a type within each patient case, by utilizing the domain knowledge of karyotyping. Evaluation results from 1909 karyotyping cases showed that the proposed Varifocal-Net achieved the highest accuracy per patient case (%) of 99.2 for both type and polarity tasks. It outperformed state-of-the-art methods, demonstrating the effectiveness of our varifocal mechanism, multi-scale feature ensemble, and dispatch strategy. The proposed method has been applied to assist practical karyotype diagnosis. Copyright © 2018, The Authors. All rights reserved.

关键词： Computer aided diagnosis

image Stitching with single-hidden layer feedforward Neural Networks

学校读者我要写书评

暂无评论

Image Stitching with single-hidden layer feedforward Neural ...

International Joint Conference on Neural Networks

作者： Min Yan Qian Yin Ping Guo Image Processing and Pattern Recognition Laboratory Beijing Normal University

ISBN: (纸本)9781509006212

In this paper, a novel image stitching method is proposed, which utilizes scale-invariant feature transform (SIFT) feature and single-hidden layer feedforward neural network (SLFN) to get higher precision of parameter estimation. In this method, features are extracted from the image sets by the SIFT descriptor and form into the input vector of the SLFN. The output of the SLFN is those translation, rotation and scaling parameters with respect to reference and registered image sets. We also apply a fast learning scheme, called pseudoinverse learning, to train SLFN to get higher training efficiency. Comparative experiments are performed between our proposed method and the traditional random sample consensus (RANSAC) based method. The results show that our method has the advantage not only at accuracy but also remarkably at fast speed.

关键词： SIFT descriptor Feedforward neural network image registration image stitching feedforward neural nets image stitching image registration Parameter estimation fast speed

Kernel Selection with Evolutionary Algorithm for Multiple Kernel Independent Component Analysis

学校读者我要写书评

暂无评论

Kernel Selection with Evolutionary Algorithm for Multiple Ke...

International Joint Conference on Neural Networks

作者： Peng Wu Qian Yin Ping Guo Image Processing and Pattern Recognition Laboratory Beijing Normal University

ISBN: (纸本)9781509006212

Kernel independent component analysis (KICA) has an important application in blind source separation, in which how to select the optimal kernel, including the kernel functional form and its parameters, is the key issue for obtaining the optimal performance. In practices, a single kernel is usually chosen as the kernel model of KICA in light of experience. However, selecting a suitable kernel model is a more difficult problem if one has not sufficient experience. To deal with this problem, an evolution based method to select the kernel model of KICA is proposed in this paper. There are two main features of the proposed method: one is that using a multiple kernel model, a convex combination of several single kernels, replaces the single kernel model;another is that particle swarm optimization (PSO) algorithm is utilized to find the combination weights of the composite kernel. Experiments conducted on separating one-dimensional mixed signals, nature images, and spectroscopic CCD images showed that using multiple kernels model with PSO kernel selection algorithm can enhance the performance of KICA.

关键词： Seeds Kernel evolutionary algorithm Independent component analysis

Study of point spread function of astronomical object imaging

学校读者我要写书评

暂无评论

International Conference on Information Science and Applications, ICISA 2016

作者： Yu, Jian Yin, Qian Guo, Ping College of Computer and Information Engineering Hanshan Normal University ChaozhouGuangdong521041 China Image Processing and Pattern Recognition Laboratory Beijing Normal University Beijing100875 China

ISBN: (纸本)9789811005565

The research of point spread function (PSF) of astronomical object imaging is very important to the astronomical image restoration. In this paper, the simulated atmospheric turbulent phase screen, the short exposure PSF and longexposure PSF during the astronomical object imaging are studied. One novel Gaussian process move-and-superposition model for representing the final astronomical object long exposure PSF formation is proposed. The proposed model obtains more accurate final PSF of astronomical object imaging. © Springer Science+Business Media Singapore 2016.

关键词： Optical transfer function

Classification of COPD with Multiple Instance Learning

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Cheplygina, Veronika Sørensen, Lauge Tax, David M.J. Pedersen, Jesper Holst Loog, Marco de Bruijne, Marleen Pattern Recognition Laboratory Delft University of Technology Delft Netherlands Image Group Department of Computer Science University of Copenhagen Copenhagen Denmark Department of Thoracic Surgery Rigshospitalet University of Copenhagen Copenhagen Denmark Biomedical Imaging Group Rotterdam Erasmus MC Rotterdam Netherlands

Nonconvex penalties with analytical solutions for one-bit compressive sensing

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Huang, Xiaolin Yan, Ming Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University MOE Key Laboratory of System Control and Information Processing Shanghai200240 China Department of Computational Mathematics Science and Engineering Department of Mathematics Michigan State University East LansingMI48824 United States

One-bit measurements widely exist in the real world and can be used to recover sparse signals. This task is known as one-bit compressive sensing (1bit-CS). In this paper, we propose novel algorithms based on both convex and nonconvex sparsity-inducing penalties for robust 1bit-CS. We consider the dual problem, which has only one variable and provides a sufficient condition to verify whether a solution is globally optimal or not. For positive homogeneous penalties, a globally optimal solution can be obtained in two steps: a proximal operator and a normalization step. For other penalties, we solve the dual problem, and it needs to evaluate the proximal operators for many times. Then we provide fast algorithms for finding analytical solutions for three penalties: minimax concave penalty (MCP), 0norm, and sorted 1penalty. Specifically, our algorithm is more than 200 times faster than the existing algorithm for MCP. Its efficiency is comparable to the algorithm for the 1penalty in time, while its performance is much better than 1. Among these penalties, sorted 1is most robust to noise in different settings. Copyright © 2017, The Authors. All rights reserved.

关键词： Compressed sensing