检索结果-内蒙古大学图书馆

Learning Counterfactual Explanation of Graph Neural Networks via Generative Flow Network

学校读者我要写书评

暂无评论

IEEE Transactions on Artificial intelligence 2024年第9期5卷 4607-4619页

作者： He, Kangjia Liu, Li Zhang, Youmin Wang, Ye Liu, Qun Wang, Guoyin Chongqing University of Posts and Telecommunications Chongqing Key Laboratory of Computational Intelligence Key Laboratory of Big Data Intelligent Computing Chongqing400065 China

Counterfactual subgraphs explain graph neural networks (GNNs) by answering the question: 'How would the prediction change if a certain subgraph were absent in the input instance?' The differentiable proxy adjacency matrix is prevalent in current counterfactual subgraph discovery studies due to its ability to avoid exhaustive edge searching. However, a prediction gap exists when feeding the proxy matrix with continuous values and the thresholded discrete adjacency matrix to GNNs, compromising the optimization of the subgraph generator. Furthermore, the end-to-end learning schema adopted in the subgraph generator limits the diversity of counterfactual subgraphs. To this end, we propose CF-GFNExplainer, a flow-based approach for learning counterfactual subgraphs. CF-GFNExplainer employs a policy network with a discrete edge removal schema to construct counterfactual subgraph generation trajectories. Additionally, we introduce a loss function designed to guide CF-GFNExplainer's optimization. The discrete adjacency matrix generated in each trajectory eliminates the prediction gap, enhancing the validity of the learned subgraphs. Furthermore, the multitrajectories sampling strategy adopted in CF-GFNExplainer results in diverse counterfactual subgraphs. Extensive experiments conducted on synthetic and real-world datasets demonstrate the effectiveness of the proposed method in terms of validity and diversity. © 2020 IEEE.

关键词： Trajectories

Fractal autoencoder with redundancy regularization for unsupervised feature selection

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2025年第2期68卷 89-102页

作者： Meiting SUN Fangyu LI Honggui HAN School of Information Science and Technology Beijing University of Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Engineering Research Center of Digital Community Ministry of Education Beijing Institute of Artificial Intelligence

Feature selection is a crucial step in data preprocessing because feature selection reduces the dimensionality of data by eliminating irrelevant and redundant features. Since manual labeling is expensive, unsupervised feature selection has received increasing attention in recent years. However, existing unsupervised feature selection methods tend to prioritize selecting highly correlated features over exploring feature diversity. Thus, a regularized fractal autoencoder(RFAE) method is proposed to select informative features in an unsupervised way. Specifically, the fractal autoencoder network extends autoencoders to construct a correspondence neural network and a selection neural network. The correspondence neural network exploits interfeature correlations and the selection neural network selects the informative features. A redundancy regularization strategy consists of a redundancy elimination regularization term based on the dependency between features and a sparse regularization term based on the group lasso. The redundancy regularization strategy eliminates feature subset redundancy and enhances network generalization ability. Extensive experimental results on six publicly available datasets show that the proposed RFAE outperforms the compared methods regarding clustering accuracy and classification accuracy. Moreover, the proposed RFAE achieves acceptable computation efficiency.

关键词： unsupervised feature selection fractal autoencoder correspondence neural network selection neural network redundancy regularization strategy

Learning Class-Aware Local Representations for Few-Shot Remote Sensing Scene Classification

学校读者我要写书评

暂无评论

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2025年 18卷 13225-13237页

作者： Wang, Liu Zhuo, Li Zhang, Hui Li, Jiafeng Beijing University of Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing100124 China

Few-Shot Remote Sensing Scene Classification (FSRSSC) aims at recognizing novel categories with only a few labeled samples. Local representations can retain richer details than image-level feature vectors and adapt well to the characteristics of spatially scattered objects in remote sensing images. However, the existing local representation-based few-shot learning works generally overlook the role of class-semantic information. To address this issue, in this paper, a Class-aware Local Representation Learning (CLRL) method is proposed for FSRSSC. It generates class-aware support local representations by calibrating the local features with class-level global features to alleviate the problem of ignoring the class-semantic information. The Image-to-Class (I2C) metric is used to measure the similarities between query samples and support samples. Moreover, a novel Local Representation-based Self-supervised Learning (LRSL) strategy is proposed. The main idea of LRSL is to maximize the relationship consistency between the query local representations and different views of support local representations. A novel consistency loss is further proposed as the objective function of LRSL to supervise the model training. Different from the existing image-level feature vector based self-supervised learning tasks, the proposed LRSL forces the model to capture the structure information of local representations. A robust model can be trained by jointly optimizing LRSL tasks and FSRSSC tasks. The experimental results on the NWPU-RESISC45, WHU-RS19 and UC Merced datasets show the superiority of our proposed method. © 2008-2012 IEEE.

关键词： Self-supervised learning

Hierarchical Learning-Based Integrated Robust Optimal Control for Nonlinear systems

学校读者我要写书评

暂无评论

IEEE Transactions on systems, Man, and Cybernetics: systems 2025年第5期55卷 3119-3129页

作者： Zhang, Jiacheng Wang, Jingjing Han, Honggui Hou, Ying Huang, Yanting Beijing University of Technology Faculty of Information Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing100124 China

The optimal control of nonlinear systems is crucial to improve system performance. However, the uncertainties of cost functions and systems dynamics make it difficult to solve the optimal control laws. To cope with this problem, a hierarchical learning-based integrated robust optimal control (HL-IROC) method is proposed in this article. The merits of the proposed HL-IROC method are three aspects: First, a hierarchical learning-based integrated optimal control (HL-IOC) scheme, contains a system dynamic learning layer and a cost function learning layer, is designed to transform the original optimal control problem into an integrated optimization problem with control laws as decision variables. Then, the relationships between cost functions and control laws are captured to overcome the difficulties brought by uncertainties in the optimal control process. Second, a global-local cooperative robust evolutionary optimization (GL-CREO) algorithm is proposed to obtain the optimal control laws. Then, a global-local robust searching strategy is employed to deal with two types of uncertainties for improving the robustness of control laws. Third, the convergence analysis of HL-IOC and GL-CREO is discussed in theory. In the experiments, the effectiveness of HL-IROC is illustrated with a nonlinear system and a wastewater treatment process. © 2013 IEEE.

关键词： Optimal control systems

Position-aware pushing and grasping synergy with deep reinforcement learning in clutter

学校读者我要写书评

暂无评论

CAAI Transactions on intelligence Technology 2024年第3期9卷 738-755页

作者： Min Zhao Guoyu Zuo Shuangyue Yu Daoxiong Gong Zihao Wang Ouattara Sie Intelligent Robotics Laboratory Faculty of Information TechnologyBeijing University of TechnologyBeijingChina Beijing Key Laboratory of Computational Intelligence and Intelligent Systems BeijingChina Laboratory of Biomechatronics and Intelligent Robotics(BIRO) Department of Mechanical and Aerospace EngineeringNorth Carolina State UniversityRaleighNorth CarolinaUSA

The positional information of objects is crucial to enable robots to perform grasping and pushing manipulations in *** effectively perform grasping and pushing manipu-lations,robots need to perceive the position information of objects,including the co-ordinates and spatial relationship between objects(e.g.,proximity,adjacency).The authors propose an end-to-end position-aware deep Q-learning framework to achieve efficient collaborative pushing and grasping in ***,a pair of conjugate pushing and grasping attention modules are proposed to capture the position information of objects and generate high-quality affordance maps of operating positions with features of pushing and grasping *** addition,the authors propose an object isolation metric and clutter metric based on instance segmentation to measure the spatial re-lationships between objects in cluttered *** further enhance the perception capacity of position information of the objects,the authors associate the change in the object isolation metric and clutter metric in cluttered environment before and after performing the action with reward function.A series of experiments are carried out in simulation and real-world which indicate that the method improves sample efficiency,task completion rate,grasping success rate and action efficiency compared to state-of-the-art end-to-end *** that the authors’system can be robustly applied to real-world use and extended to novel *** material is available at https://***/NhG\_k5v3NnM}{https://***/NhG\_k5v3NnM.

关键词： deep learning deep neural networks intelligent robots

Fuzzy Neural Network-Based Adaptive Asymmetric Constraint Control in Wastewater Treatment Process

学校读者我要写书评

暂无评论

IEEE Transactions on Artificial intelligence 2024年第6期5卷 3284-3296页

作者： Chen, Dingyuan Yang, Cuili Qiao, Junfei Beijing University of Technology Faculty of Information Technology Beijing Laboratory for Intelligent Environmental Protection Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing Institute of Artificial Intelligence Beijing100124 China

Wastewater treatment process (WWTP) is an important mean to prevent water pollution and improve ecological environment. Dissolved oxygen (DO) and nitrate nitrogen (NO3-N) concentrations are the main indicators to affect the effluent quality (EQ). In order to achieve high accuracy control, the artificial intelligence based asymmetric constraint control method with actuator saturation processing technology is proposed in WWTP. First, the fuzzy neural network (FNN) model is used to estimate the unknown situations, in which the maximum correlation entropy criterion is introduced into the adjustment of model structure to deal with the dynamic changes of WWTP. Second, the unified actuator saturation processing model is established to achieve the stable tracking performance. Third, the asymmetric barrier Lyapunov function (BLF) is introduced into the controller design. Not only the DO and NO3-N concentrations, but also the tracking error are kept within the asymmetric constraint range to guarantee the control performance. Finally, the effectiveness of the proposed method is verified via the dynamical and constant set-points simulation experiments of benchmark simulation model 1 (BSM1). © 2020 IEEE.

关键词： Wastewater treatment

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

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

Prototype Enhancement-Based Incremental Evolution Learning for Urban Garbage Classification

学校读者我要写书评

暂无评论

IEEE Transactions on Artificial intelligence 2024年第1期5卷 398-411页

作者： Han, Honggui Fan, Xiaoye Li, Fangyu Beijing University of Technology Faculty of Information Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing Institute of Artificial Intelligence Beijing100124 China

Deep neural network (DNN) based on incremental learning provides support for efficient garbage classification tasks. However, it is always challenging to accurately learn and preserve the information of known classes for updating DNN while new tasks are continuously emerging, which also affects the generalization performance of the model. To solve these issues, an incremental evolution learning (IEL) method based on prototype enhancement is proposed to accurately preserve data and improve the model generalization ability. First, a prototype enhancement method based on multidimensional Gaussian kernel density estimation is designed, which extends the prototype of each sample based on high-dimensional nonlinear data distribution. Then, the prototype enhancement accurately represents the known class data. Second, a contrastive feature method is proposed to constrain the consistency of features between tasks, which reduces the deviation between different tasks. Then, the extraction preference caused by the class sample imbalance is balanced and the generalization ability is improved. Third, the proposed IEL is used for garbage classification with imbalanced class samples. IEL implements garbage classification by effectively adapting to the differences between known classes and new classes. Finally, experiments on four standard datasets and one public garbage dataset verify that IEL has a strong classification ability in the learning tasks with the increasing number of classes. © 2020 IEEE.

关键词： Deep neural networks

Online Aware Synapse Weighted Autoencoder for Recovering Random Missing Data in Wastewater Treatment Process

学校读者我要写书评

暂无评论

IEEE Transactions on Artificial intelligence 2024年第2期5卷 578-589页

作者： Han, Honggui Sun, Meiting Li, Fangyu Faculty of Information Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Engineering Research Center of Digital Community Ministry of Education Beijing University of Technology Beijing100124 China Faculty of Information Technology Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing University of Technology Beijing100124 China

Missing values in wastewater treatment process (WWTP) data hinder the monitoring and prediction of operational status. Therefore, various online imputation methods have been proposed to recover missing values from streaming data collected from WWTP from real time. However, existing methods tend to ignore previous learned knowledge. In this article, an online aware synapse weighted autoencoder imputation method (OASI) is proposed to impute random missing values. First, an online stacked autoencoder (OSAE) framework is constructed to capture the nonlinear structure of the recently collected data. The framework decreases the computational and storage consumption of the model training. Second, an aware synapses weighted parameter regularization strategy is presented to guide the update of model parameters and alleviate the forgetting of historical information in an online continual setup. In this way, the learned features offer a more comprehensive representation of the overall information and help enhance imputation performance. Third, two real WWTP datasets with strong nonstationarity, high-noise level and high-dimensionality are used to validate the performance of the proposed OASI. Experimental results show that the proposed OASI achieves superior performances over the existing methods even in the presence of random missing values with different missing ratios, and only costs a short running time. © 2023 IEEE.

关键词： Real time systems