检索结果-内蒙古大学图书馆

A Global Reweighting Approach for Cross-Domain Semantic Segmentation

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Zhang, Yuhang Tian, Shishun Liao, Muxin Hua, Guoguang Zou, Wenbin Xu, Chen Shenzhen Key Laboratory of Advanced Machine Learning and Applications Shenzhen University Shenzhen518060 China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen518060 China Institute of Artificial Intelligence and Advanced Communication Shenzhen University Shenzhen518060 China College of Electronics and Information Engineering Shenzhen University Shenzhen518060 China College of Mathematics and Statistics Shenzhen University Shenzhen518060 China

Unsupervised domain adaptation semantic segmentation attracts much research attention due to the expensive pixel-level annotation cost. Since the adaptation difficulty of samples is different, the weight of samples should be set independently, which is called reweighting. However, existing reweighting methods only calculate local reweighting information from predicted results or context information in batch images of two domains, which may lead to over-alignment or under-alignment problems. To handle this issue, we propose a global reweighting approach. Specifically, we first define the target centroid distance, which describes the distance between the source batch data and the target centroid. Then, we introduce a Fréchet Inception Distance metric to evaluate the domain divergence and embed it into the target centroid distance. Finally, a global reweighting strategy is proposed to enhance the knowledge transferability in the source domain supervision. Extensive experiments demonstrate that our approach achieves competitive performance and helps to improve performance in other methods. © 2022, The Authors. All rights reserved.

关键词： Semantic Segmentation

Convergence Rates of Subgradient Methods for Quasi-convex Optimization Problems

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Hu, Yaohua Li, Jiawen Yu, Carisa Kwok Wai Shenzhen Key Laboratory of Advanced Machine Learning and Applications College of Mathematics and Statistics Shenzhen University Shenzhen518060 China College of Mathematics and Statistics Shenzhen University Shenzhen518060 China Department of Mathematics and Statistics Hang Seng University of Hong Kong Hong Kong

Quasi-convex optimization acts a pivotal part in many fields including economics and finance;the subgradient method is an effective iterative algorithm for solving large-scale quasi-convex optimization problems. In this paper, we investigate the iteration complexity and convergence rates of various subgradient methods for solving quasi-convex optimization in a unified framework. In particular, we consider a sequence satisfying a general (inexact) basic inequality, and investigate the global convergence theorem and the iteration complexity when using the constant, diminishing or dynamic stepsize rules. More important, we establish the linear (or sublinear) convergence rates of the sequence under an additional assumption of weak sharp minima of Hölderian order and upper bounded noise. These convergence theorems are applied to establish the iteration complexity and convergence rates of several subgradient methods, including the standard/inexact/conditional subgradient methods, for solving quasi-convex optimization problems under the assumptions of the Hölder condition and/or the weak sharp minima of Hölderian *** Codes 65K05, 90C26 Copyright © 2019, The Authors. All rights reserved.

关键词： Convex optimization

Video salient object detection using a virtual border and guided filter

学校读者我要写书评

暂无评论

Pattern Recognition 2020年 97卷

作者： Wang, Qiong Zhang, Lu Zou, Wenbin Kpalma, Kidiyo College of Computer Science and Technology Zhejiang University of Technology No.288 Road Liuhe Hangzhou310023 China - UMR 6164 RennesF-35000 France College of Electronic and Information Engineering Shenzhen Key Laboratory of Advanced Machine Learning and Applications Guangdong Key Laboratory of Intelligent Information Processing Shenzhen University Shenzhen518060 China

In this paper, we present a novel method for salient object detection in videos. Salient object detection methods based on background prior may miss salient region when the salient object touches the frame borders. To solve this problem, we propose to detect the whole salient object via the adjunction of virtual borders. A guided filter is then applied on the temporal output to integrate the spatial edge information for a better detection of the salient object edges. At last, a global spatio-temporal saliency map is obtained by combining the spatial saliency map and the temporal saliency map together according to the entropy. The proposed method is assessed on three popular datasets (Fukuchi, FBMS and VOS) and compared to several state-of-the-art methods. The experimental results show that the proposed approach outperforms the tested methods. © 2019

关键词： Object detection

A Multi-Scale Neighborhood Feature Interaction Network for Photovoltaic Cell Defect Detection

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Liu, Yu Chen Hua, Qiang Chen, Lin Lin Dong, Chun Ru Zhang, Feng Zhang, Yong Hebei Key Laboratory of Machine Learning and Computational Intelligence Hebei University Baoding China Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences Shenzhen China

Photovoltaic power generation is a key component of the industrial field, whose power generation efficiency is affected by surface defects of photovoltaic cells. The recent success in defect detection is largely attributed to the widespread application of deep learning detection models. However, because of the various sizes of defects on the photovoltaic cells, it is difficult to effectively detect multi-scale defects, especially small defects. Additionally, a large amount of noise information in images brings more difficulties in extracting distinguishable features. To address these issues, a novel one-stage Multi-Scale Neighborhood Feature Interaction Network (MNFI-Net) is proposed in this study to identify defects of various sizes in complex backgrounds. The MNFI-Net consists of the following components: 1) a Ghost Cross Stage Module (GCSM) that is devised to reduce redundant information;2) a Neighborhood Feature Interaction Model (NFIM), which is used to improve the capability of detecting defects with various size;3) a Global Attention Mechanism (GAM) that is devised to emphasize the key features of the fused feature maps. Furthermore, we introduce a novel Balanced Efficient (BE) loss function, which integrates the efficient iou loss and normalized weighted distance loss. Extensive comparison experiments and ablation studies on public photovoltaic electroluminescence images dataset are performed and the experimental results show that the defect classification and detection results of MNFI-Net in photovoltaic cell images achieve 94.0% precision and 95.5% mAP, outperforming the state-of-the-art methods. The code and proposed models in this study can be accessed at https://***/lyc686/MNFI-Net. © 2024, The Authors. All rights reserved.

关键词： Photovoltaics

A study on the uncertainty of convolutional layers in deep neural networks

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Shen, Haojing Chen, Sihong Wang, Ran Big Data Institute College of Computer Science and Software Engineering Guangdong Key Lab. of Intelligent Information Processing Shenzhen University ShenzhenGuangdong518060 China College of Mathematics and Statistics Shenzhen University Shenzhen518060 China Shenzhen Key Laboratory of Advanced Machine Learning and Applications Shenzhen University Shenzhen518060 China

This paper shows a Min-Max property existing in the connection weights of the convolutional layers in a neural network structure, i.e., the LeNet. Specifically, the Min-Max property means that, during the back propagation-based training for LeNet, the weights of the convolutional layers will become far away from their centers of intervals, i.e., decreasing to their minimum or increasing to their maximum. From the perspective of uncertainty, we demonstrate that the Min-Max property corresponds to minimizing the fuzziness of the model parameters through a simplified formulation of convolution. It is experimentally confirmed that the model with the Min-Max property has a stronger adversarial robustness, thus this property can be incorporated into the design of loss function. This paper points out a changing tendency of uncertainty in the convolutional layers of LeNet structure, and gives some insights to the interpretability of convolution. © 2020, CC-BY.

关键词： Convolution

Incorporating Hidden Layer representation into Adversarial Attacks and Defences

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Shen, Haojing Chen, Sihong Wang, Ran Wang, Xizhao Big Data Institute College of Computer Science and Software Engineering Guangdong Key Lab. of Intelligent Information Processing Shenzhen University Guangdong Shenzhen518060 China The College of Mathematics and Statistics Shenzhen University Shenzhen518060 China The Shenzhen Key Laboratory of Advanced Machine Learning and Applications Shenzhen University Shenzhen518060 China

In this paper, we propose a defence strategy to improves adversarial robustness incorporating hidden layer representation. The key of this defence strategy aims to compress or filter input’s information including adversarial perturbation. And this defence strategy can be regarded as an activation function which can be applied any kind of neural networks. We also prove theoretically the effectiveness of this defense strategy under certain conditions. Besides, incorporating hidden layer representation we propose three types of adversarial attacks to generate three types of adversarial examples, respectively. The experiments show that our defence method can significantly improve the adversarial robustness of deep neural networks which achieves the state-of-the-art performance even though we do not adopt adversarial training. © 2020, CC BY.

关键词： Deep neural networks

Adversarial learning with Cost-Sensitive Classes

学校读者我要写书评

暂无评论

arXiv 2021年

It is necessary to improve the performance of some special classes or to particularly protect them from attacks in adversarial learning. This paper proposes a framework combining cost-sensitive classification and adversarial learning together to train a model that can distinguish between protected and unprotected classes, such that the protected classes are less vulnerable to adversarial examples. We find in this framework an interesting phenomenon during the training of deep neural networks, called Min-Max property, that is, the absolute values of most parameters in the convolutional layer approach zero while the absolute values of a few parameters are significantly larger becoming bigger. Based on this Min-Max property which is formulated and analyzed in a view of random distribution, we further build a new defense model against adversarial examples for adversarial robustness improvement. An advantage of the built model is that it performs better than the standard one and can combine with adversarial training to achieve an improved performance. It is experimentally confirmed that, regarding the average accuracy of all classes, our model is almost as same as the existing models when an attack does not occur and is better than the existing models when an attack occurs. Specifically, regarding the accuracy of protected classes, the proposed model is much better than the existing models when an attack occurs. © 2021, CC BY.

关键词： Deep neural networks

GELFAND-KIRILLOV DIMENSIONS and ASSOCIATED VARIETIES of HIGHEST WEIGHT MODULES

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Bai, Zhanqiang Xiao, Wei Xie, Xun School of Mathematical Sciences Soochow University Suzhou215006 China College of Mathematics and statistics Shenzhen Key Laboratory of Advanced Machine Learning and Applications Shenzhen University Guangdong Shenzhen518060 China School of Mathematics and Statistics Beijing Institute of Technology Beijing100081 China

In this paper, we present a uniform formula of Lusztig's a-functions on classical Weyl groups. Then we obtain an efficient algorithm for the Gelfand-Kirillov dimensions of simple highest weight modules of classical Lie algebras, whose highest weight is not necessarily regular or integral. To deal with type D, we prove an interesting property about domino tableaux associated with Weyl group elements by introducing an invariant, called the hollow tableau. As an application, the associated varieties of all the simple highest weight Harish-Chandra modules are explicitly determined, including the exceptional *** Codes Primary 22E47, Secondary 17B10, 20C08 Copyright © 2020, The Authors. All rights reserved.

关键词： Algebra

Simple fourier trace formulas of cubic level and applications

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Pi, Qinghua Wang, Yingnan Zhang, Lei School of Mathematics and Statistics Shandong Univeristy Weihai Weihai264209 China Shenzhen Key Laboratory of Advanced Machine Learning and Applications College of Mathematics and Statistics Shenzhen University Shenzhen Guangdong518060 China Department of Mathematics National University of Singapore Singapore119076 Singapore

With the method of the relative trace formula and the classification of simple supercuspidal representations, we establish a simple Kuznetsov trace formula and a simple Petersson trace formula for automorphic forms on GL(2) of cubic level. As applications, we obtain some Weyl's laws for Maass newforms and a non-vanishing result for central L-values of holomorphic newforms. Copyright © 2019, The Authors. All rights reserved.

关键词：