检索结果-内蒙古大学图书馆

Representative graph neural network

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Yu, Changqian Gao, Changxin Sang, Nong Yu, Changqian Liu, Yifan Shen, Chunhua Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science & Technology China University of Adelaide Australia

Non-local operation is widely explored to model the long-range dependencies. However, the redundant computation in this operation leads to a prohibitive complexity. In this paper, we present a Representative Graph (RepGraph) layer to dynamically sample a few representative features, which dramatically reduces redundancy. Instead of propagating the messages from all positions, our RepGraph layer computes the response of one node merely with a few representative nodes. The locations of representative nodes come from a learned spatial offset matrix. The RepGraph layer is flexible to integrate into many visual architectures and combine with other operations. With the application of semantic segmentation, without any bells and whistles, our RepGraph network can compete or perform favourably against the state-of-the-art methods on three challenging benchmarks: ADE20K, Cityscapes, and PASCAL-Context datasets. In the task of object detection, our RepGraph layer can also improve the performance on the COCO dataset compared to the non-local operation. Code is available at https://***/RepGraph. Copyright © 2020, The Authors. All rights reserved.

关键词： Semantic Segmentation

Empirical studies on the properties of linear regions in deep neural networks

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zhang, Xiao Wu, Dongrui Ministry of Education Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China

A deep neural network (DNN) with piecewise linear activations can partition the input space into numerous small linear regions, where different linear functions are fitted. It is believed that the number of these regions represents the expressivity of the DNN. This paper provides a novel and meticulous perspective to look into DNNs: Instead of just counting the number of the linear regions, we study their local properties, such as the inspheres, the directions of the corresponding hyperplanes, the decision boundaries, and the relevance of the surrounding regions. We empirically observed that different optimization techniques lead to completely different linear regions, even though they result in similar classification accuracies. We hope our study can inspire the design of novel optimization techniques, and help discover and analyze the behaviors of DNNs. Copyright © 2020, The Authors. All rights reserved.

关键词： Deep neural networks

Integrating Informativeness, Representativeness and Diversity in Pool-Based Sequential Active Learning for Regression

学校读者我要写书评

暂无评论

Integrating Informativeness, Representativeness and Diversit...

International Joint Conference on Neural Networks (IJCNN)

作者： Ziang Liu Dongrui Wu Ministry of Education Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China

ISBN: (数字)9781728169262

ISBN: (纸本)9781728169279

In many real-world machine learning applications, unlabeled samples are easy to obtain, but it is expensive and/or time-consuming to label them. Active learning is a common approach for reducing this data labeling effort. It optimally selects the best few samples to label, so that a better machine learning model can be trained from the same number of labeled samples. This paper considers active learning for regression (ALR) problems. Three essential criteria - informativeness, representativeness, and diversity - have been proposed for ALR. However, very few approaches in the literature have considered all three of them simultaneously. We propose three new ALR approaches, with different strategies for integrating the three criteria. Extensive experiments on 12 datasets in various domains demonstrated their effectiveness.

关键词： Computational modeling Clustering algorithms Machine learning Training Labeling Predictive models image processing

Fault-Tolerant Consensus for Linear Singular Multi-Agent Systems with Dynamic Adaptive Event-Triggered Mechanism

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Jiang, Ping Chen, Jiejie Zeng, Zhigang Chen, Boshan The School of Computer Hubei PolyTechnic University Huangshi435000 China The School of Automation Huazhong University of Science and Technology Wuhan430074 China The College of Computer Science and Technology Hubei Normal University Huangshi435002 China The College of Mathematics and Statistics Hubei Normal University Huangshi435002 China The Key Laboratory of Image Processing and Intelligent Control Education Ministry of China Wuhan430074 China

The fault-tolerant consensus of linear singular multi-agent systems (SMASs) is studied in this paper. Firstly, a general dynamic adaptive event-triggered mechanism (ETM) is proposed, and its special cases include the existing dynamic ETM and static ETM. Then, based on this dynamic ETM, a new distributed dynamic adaptive event-triggered fault-tolerant controller (DAETFTC) is designed. Then the consensus problem is discussed for the linear SMASs via the proposed DAETFTC. Moreover, we establish novel sufficient conditions to ensure that all agents of the linear SMAS can reach consensus asymptotically. In addition, it is prove that Zeno behavior can be strictly excluded. Finally, the obtained results are simulated by a numerical example. © 2022, The Authors. All rights reserved.

关键词： Multi agent systems

Graph-based scale-aware network for human parsing 1

学校读者我要写书评

暂无评论

2nd Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2019

作者： Yang, Beibei Yu, Changqian Liu, Jiahui Gao, Changxin Sang, Nong Key Laboratory of Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China

ISBN: (数字)9783030317232

ISBN: (纸本)9783030317225

Recent work has made considerable progress in exploring contextual information for human parsing with the Fully Convolutional Network framework. However, there still exist two challenges: (1) inherent relative relationships between parts;(2) scale variation of human parts. To tackle both problems, we propose a Graph-Based Scale-Aware Network for human parsing. First, we embed a Graph-Based Part Reasoning Layer into the backbone network to reason the relative relationship between human parts. Then we construct a Scale-Aware Context Embedding Layer, which consists of two branches to capture scale-specific contextual information, with different receptive fields and scale-specific supervisions. In addition, we adopt an edge supervision to further improve the performance. Extensive experimental evaluations demonstrate that the proposed model performs favorably against the state-of-the-art human parsing methods. More specifically, our algorithm achieves 53.32% (mIoU) on the LIP dataset. © Springer Nature Switzerland AG 2019.

关键词： Embeddings

One-shot distributed algorithm for PCA with RBF kernels

学校读者我要写书评

暂无评论

arXiv 2020年

作者： He, Fan Lv, Kexin Yang, Jie Huang, Xiaolin The Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University China The MOE Key Laboratory of System Control and Information Processing 800 Dongchuan Road Shanghai200240 China

This letter proposes a one-shot algorithm for feature-distributed kernel PCA. Our algorithm is inspired by the dual relationship between sample-distributed and feature-distributed scenario. This interesting relationship makes it possible to establish distributed kernel PCA for feature-distributed cases from ideas in distributed PCA in sample-distributed scenario. In theoretical part, we analyze the approximation error for both linear and RBF kernels. The result suggests that when eigenvalues decay fast, the proposed algorithm gives high quality results with low communication cost. This result is also verified by numerical experiments, showing the effectiveness of our algorithm in practice. Copyright © 2020, The Authors. All rights reserved.

关键词： Principal component analysis

Down-Level control Based on Level Prediction of Landslide Evolutionary State

学校读者我要写书评

暂无评论

Down-Level Control Based on Level Prediction of Landslide Ev...

International Conference on Information Science and Technology (ICIST)

作者： Shu Sun Cheng Lian Xiaoping Wang School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China School of Automation Wuhan University of Technology Wuhan China School of Artificial Intelligence and Automation Key Laboratory of Image Processing and Intelligent Control Huazhong University of Science and Technology Wuhan China

Accurate prediction and effective control are the key to reducing the impact of landslide disasters. In practice, most of recent studies focus on landslide displacement prediction and control. In this paper, we propose a new control method based on the level prediction of landslide evolution state, namely down-level control. The core components are level predictor and interval predictor in this method. Specially, we first transform the traditional displacement regression prediction problem into a level classification prediction problem by labeling discrete category information for landslide sample points. Then the level predictor based on Multi-task learning-Stacked long-short time memory network (MTL-SLSTM) is established to predict the state of the landslide and judge whether the system needs to be controlled. Finally, we design a safe rainfall interval predictor based on bootstrap method to obtain the safe value of control variation. The effectiveness of the proposed control method is verified on Baishuihe and Shiliushubao landslides. The results show the proposed down-level control method is valid and more intuitive.

关键词： Information science Geology Transforms Predictive models Terrain factors Complexity theory Labeling

MBGD-RDA Training and Rule Pruning for Concise TSK Fuzzy Regression Models

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wu, Dongrui Key Laboratory of the Ministry of Education for Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China

To effectively train Takagi-Sugeno-Kang (TSK) fuzzy systems for regression problems, a Mini-Batch Gradient Descent with Regularization, DropRule, and AdaBound (MBGD-RDA) algorithm was recently proposed. It has demonstrated superior performances;however, there are also some limitations, e.g., it does not allow the user to specify the number of rules directly, and only Gaussian MFs can be used. This paper proposes two variants of MBGD-RDA to remedy these limitations, and show that they outperform the original MBGD-RDA and the classical ANFIS algorithms with the same number of rules. Furthermore, we also propose a rule pruning algorithm for TSK fuzzy systems, which can reduce the number of rules without significantly sacrificing the regression performance. Experiments showed that the rules obtained from pruning are generally better than training them from scratch directly, especially when Gaussian MFs are used. Copyright © 2020, The Authors. All rights reserved.

关键词： Gradient methods

Multi-Level temporal pyramid network for action detection

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wang, Xiang Gao, Changxin Zhang, Shiwei Sang, Nong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology China DAMO Academy Alibaba Group

Currently, one-stage frameworks have been widely applied for temporal action detection, but they still suffer from the challenge that the action instances span a wide range of time. The reason is that these one-stage detectors, e.g., Single Shot Multi-Box Detector (SSD), extract temporal features only applying a single-level layer for each head, which is not discriminative enough to perform classification and regression. In this paper, we propose a Multi-Level Temporal Pyramid Network (MLTPN) to improve the discrimination of the features. Specially, we first fuse the features from multiple layers with different temporal resolutions, to encode multi-layer temporal information. We then apply a multi-level feature pyramid architecture on the features to enhance their discriminative abilities. Finally, we design a simple yet effective feature fusion module to fuse the multi-level multi-scale features. By this means, the proposed MLTPN can learn rich and discriminative features for different action instances with different durations. We evaluate MLTPN on two challenging datasets: THUMOS14 and Activitynet v1.3, and the experimental results show that MLTPN obtains competitive performance on Activitynet v1.3 and outperforms the state-of-the-art approaches on THUMOS14 significantly. Copyright © 2020, The Authors. All rights reserved.

关键词： Feature extraction