检索结果-内蒙古大学图书馆

MetaMask: Revisiting Dimensional Confounder for Self-Supervised Learning

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Jiangmeng Qiang, Wenwen Zhang, Yanan Mo, Wenyi Zheng, Changwen Su, Bing Xiong, Hui University of Chinese Academy of Sciences Institute of Software Chinese Academy of Sciences Southern Marine Science and Engineering Guangdong Laboratory Guangzhou China Gaoling School of Artificial Intelligence Renmin University of China China Institute of Software Chinese Academy of Sciences Southern Marine Science and Engineering Guangdong Laboratory Guangzhou China Gaoling School of Artificial Intelligence Renmin University of China Beijing Key Laboratory of Big Data Management and Analysis Methods China Thrust of Artificial Intelligence The Hong Kong University of Science and Technology Guangzhou China Guangzhou HKUST Fok Ying Tung Research Institute China

As a successful approach to self-supervised learning, contrastive learning aims to learn invariant information shared among distortions of the input sample. While contrastive learning has yielded continuous advancements in sampling strategy and architecture design, it still remains two persistent defects: the interference of task-irrelevant information and sample inefficiency, which are related to the recurring existence of trivial constant solutions. From the perspective of dimensional analysis, we find out that the dimensional redundancy and dimensional confounder are the intrinsic issues behind the phenomena, and provide experimental evidence to support our viewpoint. We further propose a simple yet effective approach MetaMask, short for the dimensional Mask learned by Meta-learning, to learn representations against dimensional redundancy and confounder. MetaMask adopts the redundancy-reduction technique to tackle the dimensional redundancy issue and innovatively introduces a dimensional mask to reduce the gradient effects of specific dimensions containing the confounder, which is trained by employing a meta-learning paradigm with the objective of improving the performance of masked representations on a typical self-supervised task. We provide solid theoretical analyses to prove MetaMask can obtain tighter risk bounds for downstream classification compared to typical contrastive methods. Empirically, our method achieves state-of-the-art performance on various benchmarks. Copyright © 2022, The Authors. All rights reserved.

关键词： Redundancy

Improving Code Summarization Through Automated Quality Assurance

学校读者我要写书评

暂无评论

Improving Code Summarization Through Automated Quality Assur...

International Symposium on software Reliability engineering (ISSRE)

作者： Yuxing Hu Meng Yan Zhongxin Liu Qiuyuan Chen Bei Wang Key Laboratory of Dependable Service Computing in Cyber Physical Society (Chongqing University) Ministry of Education China School of Big Data and Software Engineering Chongqing University Chongqing China Pengcheng Laboratory Shenzhen China College of Computer Science and Technology Zhejiang University Hangzhou China

ISBN: (纸本)9781665425889

The code summarization task aims to generate brief descriptions of source code automatically. It is beneficial for developers to understand source code. However, almost all of current code summarization approaches may generate low-quality (BLEU4<40) summaries, which will mislead developers. Previous work has shown that it is possible to conduct quality assurance for document generation (QA4DG) and improve the practicability of document generation approaches. Code summarization can also be regarded as a document generation task. This work aims to investigate whether QA4DG approaches can be leveraged to improve code summarization. Specifically, we first investigate whether existing QA4DG approaches can be plugged in code summarization approaches. We find that an automated quality assurance framework for commit message generation named QACom performs best. In-spired by the idea behind QAcom, we propose an ensemble code summarization approach called Ensum. Precisely, given a code snippet, Ensum first uses current code summarization approaches to generate candidate summaries. Then, Ensum predicts the quality of each candidate summary using a collaborative filtering-based component and a retrieval-based component and selects the best candidate summary as the output. Experimental results on two public datasets show that Ensum outperforms three state-of-the-art single approaches and one ensemble approach for code summarization in terms of BLEU-4, METEOR, and ROUGE-L.

关键词： Codes Quality assurance Filtering Collaboration software reliability Task analysis

Non-Gaussian Lagrangian Stochastic Model for Wind Field Simulation in the Surface Layer

学校读者我要写书评

暂无评论

Advances in Atmospheric Sciences 2020年第1期37卷 90-104页

作者： Chao LIU Li FU Dan YANG David R.MILLER Junming WANG Key Laboratory of Dependable Service Computing in Cyber Physical Society Ministry of Education Chongqing UniversityChongqing 401331China School of Big Data and Software Engineering Chongqing UniversityChongqing 401331China Department of Natural Resources and Environment University of ConnecticutStorrsCT 06268USA Climate and Atmospheric Science Section Illinois State Water SurveyPrairie Research InstituteUniversity of Illinois at Urbana-ChampaignChampaignIL 61820USA

Wind field simulation in the surface layer is often used to manage natural resources in terms of air quality,gene flow(through pollen drift),and plant disease transmission(spore dispersion).Although Lagrangian stochastic(LS)models describe stochastic wind behaviors,such models assume that wind velocities follow Gaussian ***,measured surface-layer wind velocities show a strong skewness and *** paper presents an improved model,a non-Gaussian LS model,which incorporates controllable non-Gaussian random variables to simulate the targeted non-Gaussian velocity distribution with more accurate skewness and *** velocity statistics generated by the non-Gaussian model are evaluated by using the field data from the Cooperative Atmospheric Surface Exchange Study,October 1999 experimental dataset and comparing the data with statistics from the original Gaussian *** show that the non-Gaussian model improves the wind trajectory simulation by stably producing precise skewness and kurtosis in simulated wind velocities without sacrificing other features of the traditional Gaussian LS model,such as the accuracy in the mean and variance of simulated *** improvement also leads to better accuracy in friction velocity(i.e.,a coupling of three-dimensional velocities).The model can also accommodate various non-Gaussian wind fields and a wide range of skewness–kurtosis ***,improved skewness and kurtosis in the simulated velocity will result in a significantly different dispersion for wind/particle ***,the non-Gaussian model is worth applying to wind field simulation in the surface layer.

关键词： Lagrangian stochastic model wind field simulation non-Gaussian wind velocity surface layer

A novel method for driving path planning with spark

学校读者我要写书评

暂无评论

TechRxiv

TechRxiv 2021年

作者： Li, Leixiao Lin, Hao Wan, Jianxiong Wang, Yongsheng Gao, Jing College of Data Science and Application Inner Mongolia University of Technology China School of Computer Science and Engineering Tianjin University of Technology China Inner Mongolia Autonomous Region Engineering and Technology Research Center of Big Data Based Software Service China College of Computer and Information Engineering Inner Mongolia Agricultural University China Inner Mongolia Autonomous Region Key Laboratory of big data research and application for agriculture and animal husbandry China

Efficient and accurate driving path planning can help drivers drive. To solve the problem of low efficiency of traditional heuristic algorithms such as PSO and GA in solving driving path planning, we introduce Excellence Coefficient into heuristic algorithms and make a parallel design based on Spark, which called EC-SPPSOGA. Excellence Coefficient can increase the probability of good edges being left, simultaneously, preserves the possibility of longer side being selected. The parallel design is based on time-consuming analysis of heuristic algorithms. We validate the performance of EC-SPPSOGA based on the data in TSPLIB. It is verified that the EC-SPPSOGA can improve efficiency of driving path planning and has good scalability. © 2021, CC BY.

关键词： Heuristic algorithms

Identifying topologies and system parameters of uncertaintime-varying delayed complex networks

学校读者我要写书评

暂无评论

Science China(Technological Sciences) 2019年第1期62卷 94-105页

作者： WANG Xiong GU HaiBo WANG QianYao Lü JinHu LSC Academy of Mathematics and Systems Science Chinese Academy of Sciences School of Mathematical Sciences University of Chinese Academy of Sciences School of Automation Science and Electrical Engineering State Key Laboratory of Software Development Environment and Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University

Node dynamics and network topologies play vital roles in determining the network features and network dynamical *** it is of great theoretical significance and practical value to recover the topology structures and system parameters of uncertain complex networks with available information. This paper presents an adaptive anticipatory synchronization-based approach to identify the unknown system parameters and network topological structures of uncertain time-varying delayed complex networks in the presence of noise. Moreover, during the identification process, our proposed scheme guarantees anticipatory synchronization between the uncertain drive and constructed auxiliary response network simultaneously. Particularly, our method can be extended to several special cases. Furthermore, numerical simulations are provided to verify the effectiveness and applicability of our method for reconstructing network topologies and node parameters. We hope our method can provide basic insight into future research on addressing reconstruction issues of uncertain realistic and large-scale complex networks.

关键词： system parameters and network topologies identification anticipatory synchronization uncertain time-varying delayed complex networks noise-perturbed complex networks

Rethinking the Sample Relations for Few-Shot Classification

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Yin, Guowei Huang, Sheng Huangfu, Luwen Zhang, Yi Zhang, Xiaohong School of Big Data and Software Engineering Chongqing University No.55 Daxuecheng South Rd. Shapingba Chongqing401331 China Ministry of Education Key Laboratory of Dependable Service Computing in Cyber Physical Society Chongqing University No.174 Shazhengjie Shapingba Chongqing400044 China Fowler College of Business San Diego State University San DiegoCA92182 United States San Diego State University San DiegoCA92182 United States AI4Business Lab San Diego State University San DiegoCA92182 United States

Feature quality is paramount for classification performance, particularly in few-shot scenarios. Contrastive learning, a widely adopted technique for enhancing feature quality, leverages sample relations to extract intrinsic features that capture semantic information and has achieved remarkable success in Few-Shot Learning (FSL). Nevertheless, current few-shot contrastive learning approaches often overlook the semantic similarity discrepancies at different granularities when employing the same modeling approach for different sample relations, which limits the potential of few-shot contrastive learning. In this paper, we introduce a straightforward yet effective contrastive learning approach, Multi-Grained Relation Contrastive Learning (MGRCL), as a pretraining feature learning model to boost few-shot learning by meticulously modeling sample relations at different granularities. MGRCL categorizes sample relations into three types: intra-sample relation of the same sample under different transformations, intra-class relation of homogenous samples, and inter-class relation of inhomogeneous samples. In MGRCL, we design Transformation Consistency Learning (TCL) to ensure the rigorous semantic consistency of a sample under different transformations by aligning predictions of input pairs. Furthermore, to preserve discriminative information, we employ Class Contrastive Learning (CCL) to ensure that a sample is always closer to its homogenous samples than its inhomogeneous ones, as homogenous samples share similar semantic content while inhomogeneous samples have different semantic content. Our method is assessed across four popular FSL benchmarks, showing that such a simple pre-training feature learning method surpasses a majority of leading FSL methods. Moreover, our method can be incorporated into other FSL methods as the pre-trained model and help them obtain significant performance gains. © 2025, CC BY-NC-SA.

关键词： Contrastive Learning

Stabilization of continuous-time Markov/semi-Markov jump linear systems via finite data-rate feedback

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wang, Jingyi Feng, Jianwen Xu, Chen Wu, Xiaoqun Lü, Jinhu College of Mathematics and Statistics Shenzhen University Shenzhen518060 China School of Mathematics and Statistics Hubei Key Laboratory of Computational Science Wuhan University Wuhan China State Key Laboratory of Software Development Environment School of Automation Science and Electrical Engineering Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University Beijing China

This paper investigates almost sure exponential stabilization of continuous-time Markov jump linear systems (MJLSs) under communication data-rate constraints by introducing sampling and quantization into the feedback control. Different from previous works, the sampling times and the jump times are independent of each other in this paper. The quantization is recursively adjusted on the sampling time, and its updating strategy does not depend on the switching in a sampling interval. In other words, the explicit value of the switching signal in a sampling interval is not necessary. The numerically testable condition is developed to ensure almost sure exponential stabilization of MJLSs under the proposed communication and control protocols. We also drop the assumption of stabilizability of all individual modes required in previous works about the switched systems. Moreover, we extend the result to the case of continuous-time semi-Markov jump linear systems (semi-MJLSs) via the semi-Markov kernel approach. Finally, some numerical examples are presented to illustrate the effectiveness on stabilization of the proposed communication and control *** Codes 93D15 © 2021, CC BY.

关键词： Markov processes

MetAug: Contrastive Learning via Meta Feature Augmentation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Li, Jiangmeng Qiang, Wenwen Zheng, Changwen Su, Bing Xiong, Hui Science & Technology on Integrated Information System Laboratory Institute of Software Chinese Academy of Sciences Beijing China University of Chinese Academy of Sciences Beijing China Guangdong China Gaoling School of Artificial Intelligence Renmin University of China Beijing China Beijing Key Laboratory of Big Data Management and Analysis Methods Beijing China Guangzhou China Department of Computer Science Engineering The Hong Kong University of Science and Technology Hong Kong

What matters for contrastive learning? We argue that contrastive learning heavily relies on informative features, or "hard" (positive or negative) features. Early works include more informative features by applying complex data augmentations and large batch size or memory bank, and recent works design elaborate sampling approaches to explore informative features. The key challenge toward exploring such features is that the source multi-view data is generated by applying random data augmentations, making it infeasible to always add useful information in the augmented data. Consequently, the informativeness of features learned from such augmented data is limited. In response, we propose to directly augment the features in latent space, thereby learning discriminative representations without a large amount of input data. We perform a meta learning technique to build the augmentation generator that updates its network parameters by considering the performance of the encoder. However, insufficient input data may lead the encoder to learn collapsed features and therefore malfunction the augmentation generator. A new margin-injected regularization is further added in the objective function to avoid the encoder learning a degenerate mapping. To contrast all features in one gradient back-propagation step, we adopt the proposed optimization-driven unified contrastive loss instead of the conventional contrastive loss. Empirically, our method achieves state-of-the-art results on several benchmark datasets. Copyright © 2022, The Authors. All rights reserved.

关键词： Backpropagation

Enhancing Pulmonary Nodule Classification performance with TriCaps-RL: A Capsule Network Reinforcement Learning Approach

学校读者我要写书评

暂无评论

Journal of Network Intelligence

Journal of Network Intelligence 2024年第3期9卷 1441-1459页

作者： Zheng, Guang-Yuan Zhang, Fu-Quan Cheng, Chen Shan, Hong-Tao Soomro, Nouman Qadeer Liu, Yi College of Information Technology Shanghai Jian Qiao University Shanghai201306 China School of Mathematics and Computer Science Yanan University Shaanxi 716000 China School of Computer and Data Science Minjiang University No.200 Xiyuangong Road Fuzhou University Town Fuzhou350108 China Digital Media Art Key Laboratory of Sichuan Province Sichuan Conservatory of Music Chengdu610021 China Fuzhou Technology Innovation Center of intelligent Manufacturing information System Minjiang University Fuzhou350108 China Fujian Province University Fuzhou350300 China College of Business Shanghai Jian Qiao University Shanghai201306 China College of Electronic and Electrical Engineering Shanghai University of Engineering Science Shanghai201620 China Software Department Mehran University of Engineering & Technology Sindh 76062 Pakistan

Distinguishing the manifestations of pulmonary nodules poses a significant challenge in the medical field, demanding the expertise of experienced radiologists. This complexity results in the high cost and inadequacy of sample labeling. In response, this study introduces a reinforcement learning approach, denoted as TriCaps-RL, for the classification of pulmonary nodules based on CT signs. This approach initially employs the Q-value loss to train a single CapsNet. When the performance of the CapsNet plateaus and further enhancement becomes challenging, the loss function is progressively replaced by a triplet metric to augment performance. Consequently, a more nuanced differentiation among various types of pulmonary nodules is achieved. The proposed method can improve its performance through interactions with radiologists during their CT image reading process, thereby mitigating the shortage of radiologists and addressing the time-consuming issue of training sample labeling in medical image research. This paper delves into the fine-grained classification of pulmonary nodules using deep reinforcement learning based on CT imaging signs. In comparison to previous studies that primarily concentrate on the benign/malignant classification of lung nodules, this fine-grained classification of nodules founded on CT imaging signs proves to be more valuable for medical practitioners in making accurate diagnostic decisions. © 2024, Taiwan Ubiquitous Information CO LTD. All rights reserved.

关键词： Lung cancer