检索结果-内蒙古大学图书馆

One-bit supervision for image classification 20

学校读者我要写书评

暂无评论

One-bit supervision for image classification

Proceedings of the 34th International Conference on Neural Information Processing Systems

作者： Hengtong Hu Lingxi Xie Zewei Du Richang Hong Qi Tian Key Laboratory of Knowledge Engineering with Big Data Hefei University of Technology and School of Computer Science and Information Engineering Hefei University of Technology Huawei Inc.

ISBN: (纸本)9781713829546

This paper presents one-bit supervision, a novel setting of learning from incomplete annotations, in the scenario of image classification. Instead of training a model upon the accurate label of each sample, our setting requires the model to query with a predicted label of each sample and learn from the answer whether the guess is correct. This provides one bit (yes or no) of information, and more importantly, annotating each sample becomes much easier than finding the accurate label from many candidate classes. There are two keys to training a model upon one-bit supervision: improving the guess accuracy and making use of incorrect guesses. For these purposes, we propose a multi-stage training paradigm which incorporates negative label suppression into an off-the-shelf semi-supervised learning algorithm. In three popular image classification benchmarks, our approach claims higher efficiency in utilizing the limited amount of annotations.

关键词：

Learning to transfer graph embeddings for inductive graph based recommendation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wu, Le Yang, Yonghui Chen, Lei Lian, Defu Hong, Richang Wang, Meng Key Laboratory of Knowledge Engineering with Big Data Hefei University of Technology School of Computer Science and Information Engineering Hefei University of Technology University of Science and Technology of China

With the increasing availability of videos, how to edit them and present the most interesting parts to users, i.e., video highlight, has become an urgent need with many broad applications. As users’ visual preferences are subjective and vary from person to person, previous generalized video highlight extraction models fail to tailor to users’ unique preferences. In this paper, we study the problem of personalized video highlight recommendation with rich visual content. By dividing each video into non-overlapping segments, we formulate the problem as a personalized segment recommendation task with many new segments in the test stage. The key challenges of this problem lie in: the cold-start users with limited video highlight records in the training data and new segments without any user ratings at the test stage. To tackle these challenges, an intuitive idea is to formulate a user-item interaction graph and perform inductive graph neural network based models for better user and item embedding learning. However, the graph embedding models fail to generalize to unseen items as these models rely on the item content feature and item link information for item embedding calculation. To this end, we propose an inductive Graph based Transfer learning framework for personalized video highlight Recommendation (TransGRec). TransGRec is composed of two parts: a graph neural network followed by an item embedding transfer network. Specifically, the graph neural network part exploits the higher-order proximity between users and segments to alleviate the user cold-start problem. The transfer network is designed to approximate the learned item embeddings from graph neural networks by taking each item’s visual content as input, in order to tackle the new segment problem in the test phase. We design two detailed implementations of the transfer learning optimization function, and we show how the two parts of TransGRec can be efficiently optimized with different transfer learning optimiz

关键词： Graph neural networks

A second-scale periodicity in an active repeating fast radio burst source

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Du, Chen Huang, Yong-Feng Geng, Jin-Jun Gao, Hao-Xuan Zhang, Li Deng, Chen Cui, Lang Liao, Jie Jiang, Peng-Fei Zhang, Liang Wang, Pei Hu, Chen-Ran Dong, Xiao-Fei Xu, Fan Li, Liang Zou, Ze-Cheng Kurban, Abdusattar School of Astronomy and Space Science Nanjing University Jiangsu Nanjing210023 China Key Laboratory of Modern Astronomy and Astrophysics Nanjing University Ministry of Education Jiangsu Nanjing210023 China Purple Mountain Observatory Chinese Academy of Sciences Jiangsu Nanjing210023 China College of Big Data and Information Engineering Guizhou University Guizhou Guiyang550025 China State Key Laboratory of Public Big Data Guizhou University Guizhou Guiyang550025 China Xinjiang Astronomical Observatory Chinese Academy of Sciences Xinjiang Urumqi830011 China Key Laboratory of Radio Astronomy and Technology Chinese Academy of Sciences Beijing100101 China Xinjiang Key Laboratory of Radio Astrophysics Xinjiang Urumqi830011 China College of Astronomy and Space Science University of Chinese Academy of Sciences Beijing101408 China Guizhou Vocational College of Economics and Business Guizhou Duyun558022 China CAS Key Laboratory of FAST NAOC Chinese Academy of Sciences Beijing100101 China Institute for Frontiers in Astronomy and Astrophysics Beijing Normal University Beijing102206 China Department of Physics Anhui Normal University Anhui Wuhu241002 China Institute of Fundamental Physics and Quantum Technology Ningbo University Zhejiang Ningbo315211 China Department of Physics School of Physical Science and Technology Ningbo University Zhejiang Ningbo315211 China INAF-Osservatorio Astronomico d’Abruzzo Teramo64100 Italy

Fast radio bursts (FRBs) are fierce radio flashes from the deep sky. Abundant observations have indicated that highly magnetized neutron stars might be involved in these energetic bursts, but the underlying trigger mechanism is still enigmatic. Especially, the widely expected periodicity connected to the spin of the central engine has never been discovered, which leads to further debates on the nature of FRBs. Here we report the first discovery of a ∼ 1.7 s period in the repeating source of FRB 20201124A. This is an active repeater, from which more than 2800 FRBs have been observed on a total of 49 days. The phase-folding method is adopted to analyze the bursts on each day separately. While no periodical signal is found in all other datasets, a clear periodicity does appear on two specific days, i.e. a period of 1.706015(2) s on MJD 59310, and a slightly larger period of 1.707972(1) s on MJD 59347. A period derivative of 6.14 × 10−10 s s−1 can be derived from these two periods, which further implies a surface magnetic field strength of 1.04 × 1015 G and a spin-down age of 44 years for the central engine. It is thus concluded that FRB 20201124A should be associated with a young magnetar. © 2025, CC BY.

关键词： Magnetoplasma

Multi-Channel Co-Attention Network for Visual Question Answering

学校读者我要写书评

暂无评论

Multi-Channel Co-Attention Network for Visual Question Answe...

International Joint Conference on Neural Networks (IJCNN)

作者： Weidong Tian Bin He Nanxun Wang Zhongqiu Zhao Key Laboratory of Knowledge Engineering with Big Data Hefei University of Technology Hefei China School of Computer Science and Information Engineering Hefei University of Technology Hefei China

ISBN: (数字)9781728169262

ISBN: (纸本)9781728169279

Visual Question Answering (VQA) is to reason out correct answers based on input questions and images. Significant progresses have been made by learning rich embedding features from images and questions by bilinear models. Attention mechanisms are widely used to focus on specific visual and textual information in VQA reasoning process. However, most state-of-the-art methods concentrate on fusing the global multi-modal features, while neglect local features. Besides, the dimension is reduced excessively (from K×2048 to 2048) in general visual attention, which causes a mass of visual information loss. In this paper, we propose a novel multi-channel co-attention network (MC-CAN), which integrates multi-modal features from global level to local level. We design different multi-channel attention mechanisms separately for visual (from K×2048 to M×2048) and textual features at different level of integrations. Additionally, we further improve our proposed approach by combining it with the complementary modules such as the MLB and the Count modules. Experiments on benchmark datasets show that our approach achieves better VQA performance than other state-of-the-art methods.

关键词： Visualization Task analysis Feature extraction Correlation knowledge discovery Computational modeling Semantics

Adaptive domain of dynamic distribution based on manifold space

学校读者我要写书评

暂无评论

Adaptive domain of dynamic distribution based on manifold sp...

IEEE International Conference on Big knowledge (ICBK)

作者： Daoyuan Yu Xuegang Hu Key Laboratory of Knowledge Engineering with Big Data (Hefei University of Technology) Ministry of Education School of Computer Science and Information Engineering Hefei University of Technology Hefei China

ISBN: (数字)9781728181561

ISBN: (纸本)9781728181578

Domain adaption aims to use the source domain knowledge to assist the model learning. Most of the existing methods are based on the feature representation learning model, which are achieved by aligning the data distribution between two domains. However, in the process of feature representation learning, due to the diversity of distribution differences between domains, it faces the problems of degenerated feature transformation and unevaluated distribution alignment, which bring challenges to the existing research. Therefore, the dynamic distribution alignment factors of marginal distribution and conditional distribution are introduced to reduce the spatial distortion in the process of feature representation, and the domain adaptation is realized based on the dynamic manifold regularization constraints and structural risk minimization learning. In this paper, we propose the structural risk minimization and an improved dynamic manifold regularization constraint to solve these problems. Experimental results show that compared with traditional methods and deep level methods, the algorithm in this paper has a significant improvement in classification accuracy.

关键词： Manifolds Feature extraction Task analysis Distortion Adaptation models Risk management Kernel

Max-Min Fairness in IRS-Aided MISO Broadcast Channel via Joint Transmit and Reflective Beamforming

学校读者我要写书评

暂无评论

Max-Min Fairness in IRS-Aided MISO Broadcast Channel via Joi...

GLOBECOM 2020 - 2020 IEEE Global Communications Conference

作者： Caihong Kai Wenqi Ding Wei Huang Key Laboratory of Knowledge Engineering with Big Data (Hefei University of Technology) School of Computer Science and Information Engineering Hefei University of Technology Ministry of Education Hefei China

ISBN: (数字)9781728182988

ISBN: (纸本)9781728182995

The potential application of intelligent reflecting surfaces (IRSs) for future wireless cellular communication systems has motivated the study of metasurface for achieving additional space degree of freedom, where IRS is used to enhance the desired signal strength and suppress the interference. In this paper, by using the additional design degree of freedom provided by the IRS, we jointly optimize the transmit beamforming vector at the BS and the reflective beamforming vector at the IRS to maximize the minimum rate in the IRS-aided multi-user multiple-input-single-output broadcast channel (MISO-BC), subject to the unit modulus constraints of the reflective beamforming vector. In order to solve the non-convex optimization problem, we propose an efficient algorithm based on alternating optimization. In particular, we optimize the transmit beamforming vectors via the second-order cone problem (SOCP) and reflective beamforming vector by using the semi-definite relaxation (SDR). Numerical results show that the use of IRS leads to significant higher SINR values than benchmark schemes without IRS.

关键词： Wireless communication Array signal processing Transmitters Interference MISO communication Optimization Signal to noise ratio

Towards efficient local causal structure learning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Yang, Shuai Wang, Hao Yu, Kui Cao, Fuyuan Wu, Xindong The School of Computer Science and Information Engineering Hefei University of Technology Hefei230601 China The School of Computer and Information Technology Shanxi University Taiyuan030006 China The Key Laboratory of Knowledge Engineering With Big Data Ministry of Education Hefei University of Technology Hefei230601 China Mininglamp Academy of Sciences Mininglamp Technology Beijing100102 China

Local causal structure learning aims to discover and distinguish direct causes (parents) and direct effects (children) of a variable of interest from data. While emerging successes have been made, existing methods need to search a large space to distinguish direct causes from direct effects of a target variable T. To tackle this issue, we propose a novel Efficient Local Causal Structure learning algorithm, named ELCS. Specifically, we first propose the concept of N-structures, then design an efficient Markov Blanket (MB) discovery subroutine to integrate MB learning with N-structures to learn the MB of T and simultaneously distinguish direct causes from direct effects of T. With the proposed MB subroutine, ELCS starts from the target variable, sequentially finds MBs of variables connected to the target variable and simultaneously constructs local causal structures over MBs until the direct causes and direct effects of the target variable have been distinguished. Using eight Bayesian networks the extensive experiments have validated that ELCS achieves better accuracy and efficiency than the state-of-the-art algorithms. Copyright © 2021, The Authors. All rights reserved.

关键词： Bayesian networks

Improving Incremental Learning: A Closer Look at the Softmax Function

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Zhai, Zheng Zhang, Jiali Wang, Haiyu Wu, Mingxin Yang, Keshun Qiao, Xiaoyan Sun, Qiang Beijing Normal University No.18 Jinfeng Road Guangdong Zhuhai519087 China Shandong Technology and Business University Shandong Yantai China Yantai Key Laboratory of Big Data Modeling and Intelligent Computing Shandong China Immersion Technology and Evaluation Shandong Engineering Research Center Shandong China School of Mathematics Sichuan University Chengdu China College of Liberal Arts and Sciences University of Illinois Urbana-Champaign IL United States Department of Statistical Sciences University of Toronto ON Canada Department of Computer Science University of Toronto ON Canada Department of Statistics and Data Science MBZUAI Abu Dhabi United Arab Emirates

This paper investigates the limitations of the widely adopted softmax cross-entropy loss in incremental learning problems. Specifically, we highlight how the shift-invariant property of this loss function can lead to multiple optimal solutions and imbalanced weights across different tasks, which exacerbates catastrophic forgetting. To address these issues, we propose two key modifications to the objective function: (1) regularizing or replacing the original joint distillation loss with a summation of task-specific distillation losses, and/or (2) utilizing a shift-sensitive prediction loss function, such as the binary cross-entropy loss. Numerical experiments validate the effectiveness of these proposals, demonstrating improved prediction accuracy and a significant reduction in forgetting. © 2024, The Authors. All rights reserved.

关键词： Adversarial machine learning

Joint item recommendation and attribute inference: An adaptive graph convolutional network approach

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wu, Le Yang, Yonghui Zhang, Kun Hong, Richang Fu, Yanjie Wang, Meng Key Laboratory of Knowledge Engineering with Big Data Hefei University of Technology School of Computer Science and Information Engineering Hefei University of Technology College of Engineering and Computer Science University of Central Florida

In many recommender systems, users and items are associated with attributes, and users show preferences to items. The attribute information describes users’ (items’) characteristics and has a wide range of applications, such as user profiling, item annotation, and feature-enhanced recommendation. As annotating user (item) attributes is a labor intensive task, the attribute values are often incomplete with many missing attribute values. Therefore, item recommendation and attribute inference have become two main tasks in these platforms. Researchers have long converged that user (item) attributes and the preference behavior are highly correlated. Some researchers proposed to leverage one kind of data for the remaining task, and showed to improve performance. Nevertheless, these models either neglected the incompleteness of user (item) attributes or regarded the correlation of the two tasks with simple models, leading to suboptimal performance of these two tasks. To this end, in this paper, we define these two tasks in an attributed user-item bipartite graph, and propose an Adaptive Graph Convolutional Network (AGCN) approach for joint item recommendation and attribute inference. The key idea of AGCN is to iteratively perform two parts: 1) Learning graph embedding parameters with previously learned approximated attribute values to facilitate two tasks;2) Sending the approximated updated attribute values back to the attributed graph for better graph embedding learning. Therefore, AGCN could adaptively adjust the graph embedding learning parameters by incorporating both the given attributes and the estimated attribute values, in order to provide weakly supervised information to refine the two tasks. Extensive experimental results on three real-world datasets clearly show the effectiveness of the proposed model. Copyright © 2020, The Authors. All rights reserved.

关键词： Collaborative filtering