检索结果-南通市图书馆

Scene text recognition via dual character counting-aware visual and semantic modeling network

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第3期67卷 313-314页

作者： Ke XIAO Anna ZHU Brian Kenji IWANA Cheng-Lin LIU School of Computer Science and artificial intelligence Wuhan University of Technology Human Interface laboratory Kyushu University State Key laboratory of Multimodal artificial intelligence Systems Institute of AutomationChinese Academy of Sciences School of artificial intelligence University of Chinese Academy of Sciences

Scene text recognition(STR) is drawing increasing attention nowadays due to its wide application in real life. Character counting information, as auxiliary information, has been shown to be effective in boosting text ... 详细信息

关键词： Decoding

Emergence of machine language: towards symbolic intelligence with neural networks

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

National Science Review 2024年第4期11卷 27-30页

作者： Yuqi Wang Xu-Yao Zhang Cheng-Lin Liu Tieniu Tan Zhaoxiang Zhang State Key laboratory of Multimodal artificial intelligence Systems Institute of AutomationChinese Academy of Sciences

Representation learning is a core issue in artificial intelligence (AI). Currently,there exists a disparity in the choice of representation between humans and machines. Humans rely on discrete language for communicati...

关键词：

A Bio-Inspired Integration Model of Basal Ganglia and Cerebellum for Motion Learning of a Musculoskeletal Robot

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Journal of Systems Science & Complexity 2024年第1期37卷 82-113页

作者： ZHANG Jinhan CHEN Jiahao ZHONG Shanlin QIAO Hong State Key laboratory of Multimodal artificial intelligence Systems Institute of AutomationChinese Academy of SciencesBeijing 100190China School of artificial intelligence University of Chinese Academy of SciencesBeijing 100049China.

It is a significant research direction for highly complex musculoskeletal robots that how to develop the ability of motion learning and *** cooperations of multiple brain regions are crucial to improving motion *** by the neural mechanisms of structures,functions,and interconnections of basal ganglia and cerebellum,a biologically inspired integration model for motor learning of musculoskeletal robots is *** on the neural characteristics of the basal ganglia,the basal ganglia actor network,which mainly simulates the dorsal striatum,outputs motion commands,and the basal ganglia critic network,which simulates the ventral striatum,estimates actionstate *** network parameters are updated using the soft actor-critic *** on the sensorimotor prediction mechanism of the cerebellum,the cerebellum network evaluates the state feature extraction quality of the basal ganglia actor network and then updates the weights of its feature *** learning method is proven to converge to the optimal ***,drawing on the mechanism of dopaminergic dynamic regulation in the basal ganglia,the adaptive adjustment of target entropy and the dopaminergic experience replay are proposed to further improve the integration model,which contributes to the exploration-exploitation trade-off of motor *** bio-inspired integration model is validated on a musculoskeletal *** results indicate that this model can effectively control the musculoskeletal robot to accomplish the motion task from random starting locations to random target positions with high precision and robustness.

关键词： Basal ganglia and cerebellum bio-inspired integration model motion learning musculoskeletal robot reinforcement learning.

维普期刊数据库

Unified Classification and Rejection:A One-versus-all Framework

在线全文

学校读者我要写书评

暂无评论

Machine intelligence Research 2024年第5期21卷 870-887页

作者： Zhen Cheng Xu-Yao Zhang Cheng-Lin Liu State Key laboratory of Multimodal artificial intelligence Systems Institute of AutomationChinese Academy of SciencesBeijing100190China School of artificial intelligence University of Chinese Academy of SciencesBeijing100049China

Classifying patterns of known classes and rejecting ambiguous and novel(also called as out-of-distribution(OOD))inputs are involved in open world pattern *** neural network models usually excel in closed-set classification while perform poorly in rejecting OOD *** tackle this problem,numerous methods have been designed to perform open set recognition(OSR)or OOD rejection/detection *** methods mostly take post-training score transformation or hybrid models to ensure low scores on OOD inputs while separating known *** this paper,we attempt to build a unified framework for building open set classifiers for both classification and OOD *** formulate the open set recognition of K-known-class as a(K+1)-class classification problem with model trained on known-class samples *** decomposing the K-class problem into K one-versus-all(OVA)binary classification tasks and binding some parameters,we show that combining the scores of OVA classifiers can give(K+1)-class posterior probabilities,which enables classification and OOD rejection in a unified *** maintain the closed-set classification accuracy of the OVA trained classifier,we propose a hybrid training strategy combining OVA loss and multi-class cross-entropy *** implement the OVA framework and hybrid training strategy on the recently proposed convolutional prototype network and prototype classifier on vision transformer(ViT)*** on popular OSR and OOD detection datasets demonstrate that the proposed framework,using a single multi-class classifier,yields competitive performance in closed-set classification,OOD detection,and misclassification *** code is available at https://***/zhen-cheng121/CPN_OVA_unified.

关键词： Open set recognition out-of-distribution detection misclassification detection convolutional prototype network oneversus-all Dempster-Shafer theory of evidence

Adaptive 5G-and-beyond network-enabled interpretable federated learning enhanced by neuroevolution

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第7期67卷 107-132页

作者： Bin CAO Jianwei ZHAO Xin LIU Yun LI State Key laboratory of Reliability and intelligence of Electrical Equipment Hebei University of Technology School of artificial intelligence Hebei University of Technology School of Economics and Management Hebei University of Technology Industrial artificial intelligence Centre Shenzhen Institute for Advanced StudyUniversity of Electronic Science and Technology of China

Mobile telemedicine systems based on the next-generation communication will significantly enhance deep fusion of network automation and federated learning(FL), but data privacy is a paramount issue in sectors like healthcare. This work hence considers FL augments 5G-and-beyond networks by training deep learning(DL) models without the need to exchange raw data. The substantial communication loads imposed on by extensive parameters involved in DL models are managed through adaptive scheduling mechanisms effectively. To address the opaque nature of DL models and to improve the interpretability of FL models, we introduce a convolutional fuzzy rough neural network specifically designed for medical image processing. We also develop a multiobjective memetic evolutionary algorithm to streamline and optimize the neural network architectures. Our comprehensive FL framework integrates smart scheduling, interpretable fuzzy rough logic,and neuroevolution. This framework is shown to improve communication efficiency, increase interpretability of diagnosis with protected privacy, and generate low-complexity neural architectures.

关键词： 5G-and-beyond network interpretable federated learning mobile telemedicine system fuzzy rough set theory neuroevolution

EmotionIC: emotional inertia and contagion-driven dependency modeling for emotion recognition in conversation

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第08期67卷 130-146页

作者： Yingjian LIU Jiang LI Xiaoping WANG Zhigang ZENG School of artificial intelligence and Automation Huazhong University of Science and Technology Hubei Key laboratory of Brain-inspired Intelligent Systems Huazhong University of Science and Technology Key laboratory of Image Processing and Intelligent Control (Huazhong University of Science and Technology) Ministry of Education Institute of artificial intelligence Huazhong University of Science and Technology

Emotion recognition in conversation(ERC) has attracted growing attention in recent years as a result of the advancement and implementation of human-computer interface technologies. In this paper,we propose an emotional inertia and contagion-driven dependency modeling approach(EmotionIC) for ERC tasks. Our EmotionIC consists of three main components, i.e., identity masked multi-head attention(IMMHA), dialogue-based gated recurrent unit(DiaGRU), and skip-chain conditional random field(SkipCRF).Compared to previous ERC models, EmotionIC can model a conversation more thoroughly at both the feature-extraction and classification levels. The proposed model attempts to integrate the advantages of attention-and recurrence-based methods at the feature-extraction level. Specifically, IMMHA is applied to capture identity-based global contextual dependencies, while Dia GRU is utilized to extract speaker-and temporal-aware local contextual information. At the classification level, SkipCRF can explicitly mine complex emotional flows from higher-order neighboring utterances in the conversation. Experimental results show that our method can significantly outperform the state-of-the-art models on four benchmark *** ablation studies confirm that our modules can effectively model emotional inertia and contagion.

关键词： emotion recognition in conversation emotional inertia and contagion multi-head attention gated recurrent unit conditional random field

同方期刊数据库

A hybrid attention model based on first-order statistical features for smoke recognition

在线全文

学校读者我要写书评

暂无评论

Science China(Technological Sciences) 2024年第3期67卷 809-822页

作者： GUO Nan LIU JiaHui DI KeXin GU Ke QIAO JunFei Beijing laboratory of Smart Environmental Protection Beijing Key Laboratory of Computational Intelligence and Intelligent SystemBeijing Artificial Intelligence InstituteBeijing 100124China Faculty of Information Technology Beijing University of TechnologyBeijing 100124China

Smoke and fire recognition are of great importance on foreseeing fire disasters and preventing environmental pollution by monitoring the burning process of objects(e.g., straw, fuels). However, since fire images suffer from problems like the variability of the features, complexity of scenarios, interference from background, changeable weather conditions as well as image quality problems, identifying smoke and fire accurately and promptly from a given image still remains a substantial challenge. Automatically learning the features of smoke images by CNNs has improved the target recognition ability compared to traditional approaches,nonetheless, convolutions and pooling operations in CNNs may cause severe information loss which may lead to *** tackle the above problems, this paper proposed a hybrid attention model based on the characteristics of smoke images. This model adopted multiple optimized attention mechanism in several stages to quickly and precisely capture the important features,achieving state-of-the-art performance on smoke and fire recognition in terms of accuracy and speed. Our proposed module mainly consists of two stages: pooling and attention. In the first stage, we conducted several newly proposed first-order pooling methods. Through traversing the data space in a larger scope, features are better reserved, thus constructing a more intact feature space of smoke and fire in an image. In the second stage, feature maps are aggregated together to perform channel and spatial attention. The channel and spatial dependencies allow us to quickly catch the important features presented in an image. By fully exploring the feature space and prominent salient features, characteristics of smoke and fire are better presented so as to obtain better smoke and fire detection results. Experiments have been conducted on public smoke detection dataset and new proposed fine-grained smoke and fire detection database. Experimental results revealed that the propose

关键词： hybrid attention first-order pooling smoke and fire detection deep convolutional neural networks

维普期刊数据库

A multiscale adaptive framework based on convolutional neural network: Application to fluid catalytic cracking product yield prediction

在线全文

学校读者我要写书评

暂无评论

Petroleum Science 2024年第4期21卷 2849-2869页

作者： Nan Liu Chun-Meng Zhu Meng-Xuan Zhang Xing-Ying Lan State Key laboratory of Heavy Oil Processing China University of Petroleum (Beijing) College of artificial intelligence China University of Petroleum (Beijing)

Since chemical processes are highly non-linear and multiscale, it is vital to deeply mine the multiscale coupling relationships embedded in the massive process data for the prediction and anomaly tracing of crucial process parameters and production indicators. While the integrated method of adaptive signal decomposition combined with time series models could effectively predict process variables, it does have limitations in capturing the high-frequency detail of the operation state when applied to complex chemical processes. In light of this, a novel Multiscale Multi-radius Multi-step Convolutional Neural Network(Msrt Net) is proposed for mining spatiotemporal multiscale information. First, the industrial data from the Fluid Catalytic Cracking(FCC) process decomposition using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN) extract the multi-energy scale information of the feature subset. Then, convolution kernels with varying stride and padding structures are established to decouple the long-period operation process information encapsulated within the multi-energy scale data. Finally, a reconciliation network is trained to reconstruct the multiscale prediction results and obtain the final output. Msrt Net is initially assessed for its capability to untangle the spatiotemporal multiscale relationships among variables in the Tennessee Eastman Process(TEP). Subsequently, the performance of Msrt Net is evaluated in predicting product yield for a 2.80 × 10～6 t/a FCC unit, taking diesel and gasoline yield as examples. In conclusion, Msrt Net can decouple and effectively extract spatiotemporal multiscale information from chemical process data and achieve a approximately reduction of 30% in prediction error compared to other time-series models. Furthermore, its robustness and transferability underscore its promising potential for broader applications.

关键词： Fluid catalytic cracking Product yield Data-driven modeling Multiscale prediction Data decomposition Convolution neural network

Nested Saturated Control of Uncertain Complex Cascade Systems Using Mixed Saturation Levels

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

IEEE/CAA Journal of Automatica Sinica 2024年第5期11卷 1163-1174页

作者： Meng Li Zhigang Zeng Wuhan National laboratory for Optoelectronics Wuhan 430074 School of artificial intelligence and Automation Huazhong University of Science and TechnologyWuhan 430074China School of artificial intelligence and Automation Huazhong University of Science and TechnologyWuhan 430074 Key laboratory of Image Processing and Intelligent Control of Education Ministry of China Wuhan 430074China

This study addresses the problem of global asymptotic stability for uncertain complex cascade systems composed of multiple integrator systems and non-strict feedforward nonlinear systems. To tackle the complexity inherent in such structures, a novel nested saturated control design is proposed that incorporates both constant saturation levels and state-dependent saturation levels. Specifically, a modified differentiable saturation function is proposed to facilitate the saturation reduction analysis of the uncertain complex cascade systems under the presence of mixed saturation levels. In addition, the design of modified differentiable saturation function will help to construct a hierarchical global convergence strategy to improve the robustness of control design scheme. Through calculation of relevant inequalities, time derivative of boundary surface and simple Lyapunov function,saturation reduction analysis and convergence analysis are carried out, and then a set of explicit parameter conditions are provided to ensure global asymptotic stability in the closed-loop systems. Finally, a simplified system of the mechanical model is presented to validate the effectiveness of the proposed method.

关键词： method saturation differentiable

维普期刊数据库