检索结果-内蒙古大学图书馆

Closer to Pre-trained Network Transfer Better

学校读者我要写书评

暂无评论

Closer to Pre-trained Network Transfer Better

IEEE Joint International Information Technology and Artificial intelligence Conference (ITAIC)

作者： Siyu Chen Wei Li School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan People’s Republic of China Image Processing and Intelligent Control Key Laboratory Education Ministry of China Wuhan People’s Republic of China

ISBN: (数字)9781728152448

ISBN: (纸本)9781728152455

In recent years, Deep Neural Network (DNN) has been widely used in the domain of computer vision, but its further development is restricted because of the lack of train samples. Fine-tuning is one of deep transfer learning methods to solve this problem. Nowadays, some existing Fine-tuning approaches assume that if the fine-tuned network could be as close to the pre-trained network as possible, the source knowledge would be fully utilized to improve the study effect. Although these approaches work pretty well, the assumptions on which they are based have not been systematically proved. Due to the complexity of DNNs, offering a reasonable and accurate explanation of the performance of the network is significant for designing new algorithms. In this paper, we design a series of experiments with or without transfer learning respectively, to characterize the relationship between network differences and network performance. The results show that there is an obvious negative correlation between them under certain circumstances. This conclusion can not only provide a more reasonable explanation for the relevant transfer learning algorithm, but also provide support and inspiration for the future work.

关键词： Knowledge engineering Correlation Transfer learning Neural networks Task analysis Information technology Knowledge transfer

Synaptic learning with augmented spikes

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Yu, Qiang Song, Shiming Ma, Chenxiang Pan, Linqiang Tan, Kay Chen Tianjin Key Laboratory of Cognitive Computing and Application College of Intelligence and Computing Tianjin University Tianjin China Key Laboratory of Image Information Processing and Intelligent Control Institute of Artificial Intelligence School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China The Department of Computer Science City University of Hong Kong Hong Kong

Traditional neuron models use analog values for information representation and computation, while all-or-nothing spikes are employed in the spiking ones. With a more brain-like processing paradigm, spiking neurons are more promising for improvements on efficiency and computational capability. They extend the computation of traditional neurons with an additional dimension of time carried by all-or-nothing spikes. Could one benefit from both the accuracy of analog values and the time-processing capability of spikes? In this paper, we introduce a concept of augmented spikes to carry complementary information with spike coefficients in addition to spike latencies. New augmented spiking neuron model and synaptic learning rules are proposed to process and learn patterns of augmented spikes. We provide systematic insight into the properties and characteristics of our methods, including classification of augmented spike patterns, learning capacity, construction of causality, feature detection, robustness and applicability to practical tasks such as acoustic and visual pattern recognition. The remarkable results highlight the effectiveness and potential merits of our methods. Importantly, our augmented approaches are versatile and can be easily generalized to other spike-based systems, contributing to a potential development for them including neuromorphic computing. Copyright © 2020, The Authors. All rights reserved.

关键词： Neurons

Deep Rank-Consistent Pyramid Model for Enhanced Crowd Counting

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Gao, Jiaqi Huang, Zhizhong Lei, Yiming Shan, Hongming Wang, James Z. Wang, Fei-Yue Zhang, Junping The Shanghai Key Laboratory of Intelligent Information Processing School of Computer Science Fudan University Shanghai200433 China The Institute of Science and Technology for Brain-inspired Intelligence MOE Frontiers Center for Brain Science Fudan University Shanghai200433 China The Shanghai Center for Brain Science and Brain-Inspired Technology Shanghai201210 China The College of Information Sciences and Technology The Pennsylvania State University University ParkPA16802 United States The State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China The Institute of Systems Engineering Macau University of Science and Technology China Qingdao Academy of Intelligent Industries Qingdao266109 China

Most conventional crowd counting methods utilize a fully-supervised learning framework to establish a mapping between scene images and crowd density maps. They usually rely on a large quantity of costly and time-intensive pixel-level annotations for training supervision. One way to mitigate the intensive labeling effort and improve counting accuracy is to leverage large amounts of unlabeled images. This is attributed to the inherent self-structural information and rank consistency within a single image, offering additional qualitative relation supervision during training. Contrary to earlier methods that utilized the rank relations at the original image level, we explore such rank-consistency relation within the latent feature spaces. This approach enables the incorporation of numerous pyramid partial orders, strengthening the model representation capability. A notable advantage is that it can also increase the utilization ratio of unlabeled samples. Specifically, we propose a Deep Rank-consistEnt pyrAmid Model (DREAM), which makes full use of rank consistency across coarse-to-fine pyramid features in latent spaces for enhanced crowd counting with massive unlabeled images. In addition, we have collected a new unlabeled crowd counting dataset, FUDAN-UCC, comprising 4,000 images for training purposes. Extensive experiments on four benchmark datasets, namely UCF-QNRF, ShanghaiTech PartA and PartB, and UCF-CC-50, show the effectiveness of our method compared with previous semi-supervised methods. The codes are available at https://***/bridgeqiqi/DREAM. Copyright © 2022, The Authors. All rights reserved.

关键词： image enhancement

A survey on negative transfer

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zhang, Wen Deng, Lingfei Zhang, Lei Wu, Dongrui the 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 the School of Microelectronics and Communication Engineering Chongqing University Chongqing400044 China

—Transfer learning (TL) utilizes data or knowledge from one or more source domains to facilitate the learning in a target domain. It is particularly useful when the target domain has very few or no labeled data, due to annotation expense, privacy concerns, etc. Unfortunately, the effectiveness of TL is not always guaranteed. Negative transfer (NT), i.e., leveraging source domain data/knowledge undesirably reduces the learning performance in the target domain, has been a long-standing and challenging problem in TL. Various approaches have been proposed in the literature to handle it. However, there does not exist a systematic survey on the formulation of NT, the factors leading to NT, and the algorithms that mitigate NT. This paper fills this gap, by first introducing the definition of NT and its factors, then reviewing about fifty representative approaches for overcoming NT, according to four categories: secure transfer, domain similarity estimation, distant transfer, and NT mitigation. NT in related fields, e.g., multi-task learning, lifelong learning, and adversarial attacks, are also discussed. Copyright © 2020, The Authors. All rights reserved.

关键词： Surveys

A visual kinematics calibration method for manipulator based on nonlinear optimization

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Peng, Gang Wang, Zhihao Yang, Jin Li, Xinde Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China IEEE senior member School of Automation South East University Nanjing China

The traditional kinematic calibration method for manipulators requires precise three-dimensional measuring instruments to measure the end pose, which is not only expensive due to the high cost of the measuring instruments but also not applicable to all manipulators. Another calibration method uses a camera, but the system error caused by the camera's parameters affects the calibration accuracy of the kinematics of the robot arm. Therefore, this paper proposes a method for calibrating the geometric parameters of a kinematic model of a manipulator based on monocular vision. Firstly, the classic Denavit-Hartenberg (D-H) modeling method is used to establish the kinematic parameters of the manipulator. Secondly, nonlinear optimization and parameter compensation are performed. The three-dimensional positions of the feature points of the calibration plate under each manipulator attitude corresponding to the actual kinematic model and the classic D-H kinematic model are mapped into the pixel coordinate system, and the sum of Euclidean distance errors of the pixel coordinates of the two is used as the objective function to be optimized. The experimental results show that the pixel deviation of the end pose corresponding to the optimized D-H kinematic model proposed in this paper and the end pose corresponding to the actual kinematic model in the pixel coordinate system is 0.99 pixels. Compared with the 7.9 deviation pixels between the pixel coordinates calculated by the classic D-H kinematic model and the actual pixel coordinates, the deviation is reduced by nearly 7 pixels for an 87% reduction in error. Therefore, the proposed method can effectively avoid system errors caused by camera parameters in visual calibration, can improve the absolute positioning accuracy of the end of the robotic arm, and has good economy and universality. Copyright © 2020, The Authors. All rights reserved.

关键词： Pixels

Observer-Based Robust Containment control of Multi-agent Systems With Input Saturation

学校读者我要写书评

暂无评论

Observer-Based Robust Containment Control of Multi-agent Sys...

第三十九届中国控制会议

作者： Juan Qian Xiaoling Wang Guo-Ping Jiang Housheng Su College of Automation and College of Artificial Intelligence Nanjing University of Posts and Telecommunicationsand Jiangsu Engineering Lab for IOT Intelligent Robots(IOTRobot) School of Artificial Intelligence and Automation Image Processing and Intelligent Control Key Laboratory of Education Ministry of China Huazhong University of Science and Technology

In this paper, the robust containment control problem of the leader-following multi-agent systems with input saturation and input additive disturbance is addressed, where the followers can be informed by multiple leaders. With the help of the lowand-high gain feedback technique and the high-gain observer approach, a distributed control algorithm for each agent is firstly designed by using the observed output information, then sufficient conditions are provided to guarantee the semi-global robust containment of the system. Finally, some numerical simulations are given to verify the correctness of the theoretical results.

关键词： Containment control multi-agent system input saturation input additive disturbance

Parity space-based model mismatch detection for linear discrete time-invariant systems with unknown disturbances

学校读者我要写书评

暂无评论

Parity space-based model mismatch detection for linear discr...

第三十九届中国控制会议

作者： Yi Tang Dan Ling Hong Zhang Yanewei Wang Ying Zheng the Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and AutomationHuazhong University of Science and Technology School of Electrical and Information Engineering Zhengzhou University of Light Industry School of Mechanical & Electrical Engineering Wuhan Institute of Technology

Model mismatch is one of the main factors of control performance degradation. In this paper, a new model mismatch detection approach with parity space-based methods is proposed for linear discrete time-invariant（LDTI） systems. This paper establishes a parity space-based residual evaluation function for the detection of model mismatch. The problem of parity space-based model mismatch is formulated on a state space model of a LDTI system. A standard evaluation function and its thresholds for the detection of model mismatch are proposed by providing the parity model-based residuals. It is shown the proposed approach results in an optimal fault detectability without any assumptions on the probability distribution of unknown disturbances. The effectiveness of the proposed methods is demonstrated by the simulation study of the paper machine head box control example.

关键词： Parity space Model mismatch detection LDTV systems model-based residual generation

Adversarial refinement network for human motion prediction

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Chao, Xianjin Bin, Yanrui Chu, Wenqing Cao, Xuan Ge, Yanhao Wang, Chengjie Li, Jilin Huang, Feiyue Leung, Howard City University of Hong Kong Hong Kong Hong Kong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China Tencent Youtu Lab Shanghai China

Human motion prediction aims to predict future 3D skeletal sequences by giving a limited human motion as inputs. Two popular methods, recurrent neural networks and feed-forward deep networks, are able to predict rough motion trend, but motion details such as limb movement may be lost. To predict more accurate future human motion, we propose an Adversarial Refinement Network (ARNet) following a simple yet effective coarse-to-fine mechanism with novel adversarial error augmentation. Specifically, we take both the historical motion sequences and coarse prediction as input of our cascaded refinement network to predict refined human motion and strengthen the refinement network with adversarial error augmentation. During training, we deliberately introduce the error distribution by learning through the adversarial mechanism among different subjects. In testing, our cascaded refinement network alleviates the prediction error from the coarse predictor resulting in a finer prediction robustly. This adversarial error augmentation provides rich error cases as input to our refinement network, leading to better generalization performance on the testing dataset. We conduct extensive experiments on three standard benchmark datasets and show that our proposed ARNet outperforms other state-of-the-art methods, especially on challenging aperiodic actions in both short-term and long-term predictions. © 2020, CC-BY.

关键词： Errors

Parameter training methods for convolutional neural networks with adaptive adjustment method based on Caputo fractional-order differences

学校读者我要写书评

暂无评论

Chaos, Solitons & Fractals 2025年 199卷

作者： Haiming Zhao Honggang Yang Jiejie Chen Ping Jiang Zhigang Zeng School of Computer and Information Engineering Hubei Normal University Huangshi 435000 China School of Automation Wuhan University of Technology Wuhan 430000 China School of Computer Hubei Polytechnic University Huangshi 435000 China School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan 430000 China Key Laboratory of Image Information Processing and Intelligent Control Ministry of Education of China Wuhan 430000 China

As deep learning technologies continue to permeate various sectors, optimization algorithms have become increasingly crucial in neural network training. This paper introduces two adaptive momentum algorithms based on Grünwald–Letnikov and Caputo fractional-order differences—Fractional Order Adagrad (FAdagrad) and Fractional Order Adam (FAdam)—to update parameters more flexibly by adjusting momentum information. Commencing from the definitions of fractional derivatives, we propose integrating fractional-order differences with gradient algorithms in convolutional neural networks (CNNs). These adaptive momentum algorithms, leveraging Grünwald–Letnikov and Caputo fractional-order differences, offer enhanced flexibility, thereby accelerating convergence. Our nonlinear parameter tuning method for CNNs demonstrates superior performance compared to traditional integer-order momentum algorithms and the standard Adam algorithm. Experimental results on the BraTS2021 dataset and CIFAR-100 dataset reveal that the proposed fractional-order optimization algorithms significantly outperform their integer-order counterparts in model optimization. They not only expedite convergence but also improve the accuracy of image recognition and segmentation.

关键词：