查看更多>>摘要：Y In this paper, an adaptive neural output feedback control (ANOFC) scheme is proposed for controlling an electrically driven robotic manipulator (EDRM) system with prescribed errors constraint by using a neural network-based adaptive observer (NNAO) and a backstepping design technique. First, the NNAO is designed to observe the unknown and unmeasured joint velocities of EDRM. Second, the prescribed performance bounds of output tracking are used to achieve the prescribed transient and steady-state performance based on barrier Lyapunov function. Then, the ANOFC scheme is derived by using the backstepping design methodology, where neural networks with adaptive update laws are utilized to approximate the unknown nonlinear functions. By using the Lyapunov stability analysis method, the observer and closed-loop control system can be proven to be stable such that all the uniformly bounded variables in the system are guaranteed and the output tracking errors remain within the specified prescribed bounds. Finally, the simulation results demonstrate that the proposed NNAO and ANOFC schemes can achieve the satisfied estimation capability and tracking effectiveness. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Natural gesticulation in-air is multifaceted, as defining a gesture's pattern/size/speed or its stroke information is difficult. The unavailability of vision-based techniques to distinguish the ground truth stroke segments and the intentional movements (self co-articulation) results in capturing everything on the trajectory. But recognizing the gesture patterns (A-Z, a-z, 0-9, 04 operators, and 29 symbols) along with their self co-articulations results in the rise of misclassification. Hence, gestures are separated into 3 sets based on their physical structures using an artificial neural network. Then, set specific pre-processing models are proposed to remove these self co-articulations. As a result, a mean error rate of 0.0112 (ground truth segment removed) and 5.63% of self co-articulations present in the gesture patterns is obtained. A relative improvement of 22% (accuracy - 94.17%) over the existing models is achieved. Then, gestures are clustered into two groups by Mamdani fuzzy interference system using prior stroke information. Using transfer learning, separate pre-trained AlexNet models were utilized to recognize the gesture patterns falling under each group with an average accuracy of 97.05% (precision - 0.9680, recall - 0.9656; F-score - 0.9668). This is a relative improvement of similar to 17% over the existing state-of-the-art models in gesture recognition. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：The rating data collected by the recommender system usually contains noise due to external factors such as human uncertainty and inconsistency. Such noise, usually modeled by a normal distribution, leads to a magic barrier (MGBR) to the recommender system. However, existing MGBR estimation approaches require a user-specified standard deviation of noise, or make strong assumptions about true ratings, or need additional information from experts or users. In this paper, we propose a Mixture of Gaussians (MoG) model without user intervention to handle this issue. First, the user uncertainties are modeled using MoG, which is a universal approximator for any continuous distribution. Second, we employ the expectation-maximization algorithm to determine the parameters of user uncertainty. Finally, the MGBR is computed by Bayesian formula with the parameters. Experimental results on four well-known datasets show that the MGBRs estimated by the new model are close to the results of the state-of-the-art algorithms. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：This paper addresses a special case of irregular bin packing problem which the irregular pieces with free rotation have to fill a large irregular stock sheet with defective regions while satisfying the special boundary constraint, i.e., the piece can protrude from the sheet so long as the key points in the piece's interior lie within the container. The objective of this problem is to maximize the number of filled pieces. To our best knowledge, the piece must be placed completely inside the sheet for all packing problem tackled by published literature. Thus, existing approaches are not good solutions to this special packing problem. To achieve the goal of automated arrangement of pieces and maximize the space utilization, the genetic algorithm and grey wolf optimization algorithm are designed to solve it. The genetic algorithm adopts the elitism strategy for maintaining the portion of the best chromosomes. A new method of updating the main controlling parameter is applied for reinforcing the exploration ability of the grey wolf optimization. These two algorithms use a vector of pieces as the solution representation, and a novel heuristic algorithm decodes it to produce a layout. The proposed heuristic algorithm divides the process of packing into two stages with the objective of satisfying constraints and achieving good space utilization of sheet. Computational experiments show that the presented methods can solve this new kind of the packing problem very well. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Currently, the most popular and effective approach to solve multivariate time series classification(MTSC) tasks is based on deep learning technology. However, the existing deep learning-based algorithms ignore the unique time characteristics of time series in the process of network training, and do not consider the features correlation in different convolutional layers. So they cannot obtain the convincing feature representation ability and result in unsatisfactory classification accuracy. To solve this problem, we propose a new time corrected residual attention network(TCRAN) which can fully extract the long-term time-dependence information to enhance the discriminative power of the network. The hallmark of TCRAN is that we employ the time residual channel attention block(TRCAB) as the basic structure, which incorporates the adaptive channel feature adjustment mechanism(AFM) and the bi-directional gated recurrent unit(Bi-GRU) into the deep residual structure to adaptively extract time-dependent features. Meanwhile, to integrate the overall dependency information between different layers, we also employ an inter-module adaptive feature adjustment mechanism(IAM) in the TCRAN. The experiments results with 15 multivariate time series datasets illustrate that the proposed TCRAN can achieve the highest average classification accuracy of 0.7276 and improve accuracy by 1.64% compared to the state-of-the-art method. All these verify the effectiveness of TCRAN. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Recently ensemble models were adopted to predict the credit risk commonly. Although they have a better performance generally, ensemble models are easy to be badly affected by imbalance classes which are a common issue in credit risk prediction. And the prediction model should be constructed according to the feature of complex distributions of financial data. However, these problems have not attracted enough attention. This paper constructs an ensemble model for imbalanced credit risk prediction and improves algorithms for the feature of financial data. The ensemble model mainly combines Synthetic Minority Over-sampling Technique Evaluation (SMOTE) and Multi-Kernel Fuzzy C-Means (MK-FCM) optimized by Particle Swarm Optimization (PSO). In the preprocessing phase, the multi-method of descending dimension is used to reduce the dimension. The improved SMOTE can make new synthetic samples more decentralized, which can not only balance the number of samples of different classes, but avoid the overfitting to some extent. In the base classifier construction phase, Fuzzy C-Means (FCM) is improved by multi-kernel function to build a new base classifier MK-FCM, which can synthetize the merits of multiple kernel functions to enhance the evaluated performance. The improved PSO, which has dynamically adjustable function, is used to optimize parameters for MK-FCM. In the empirical research, the sample data are from financial indicators of Chinese listed hospitality and tourism corporations, and the proposed model makes the comparative analysis with other relative models. The results from Matlab software show that the presented ensemble model has the best performance on imbalanced credit risk prediction. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：The prohibitive cost and low fidelity of experimental data in industry-scale thermofluid systems limit the usefulness of pure data-driven machine learning methods. Physics-informed neural networks (PINN) strive to overcome this by embedding the physics equations in the construction of the neural network loss function. In the present paper, the mixed-variable PINN methodology is applied to develop steady-state and transient surrogate models of incompressible laminar flow with heat transfer through a 2D internal domain with obstructions. Automatic spatial and temporal differentiation is applied to the partial differential equations for mass, momentum and energy conservation, and the residuals are included in the loss function, together with the boundary and initial values. Good agreement is obtained between the PINN and CFD results for both the steady-state and transient cases, but normalization of the PDEs proves to be crucial. Although this proves the ability of the PINN approach to solve multiple physics-based PDEs on a single domain, the PINN takes significantly longer to solve than the traditional finite volume numerical methods utilized in commercial CFD software. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Nowadays, designing a reliable hub network has become a critical issue in the process of transporting goods from an origin to a destination. Due to intentional disasters, both location and protection of hub play a key role in satisfying the demands and ensuring network reliability. This study tries to model the impact of the number of hubs opened, allocation of defensive resources, and the risk of disruptions on the configuration of the hub network. This model aims is to minimize the total transportation cost via the primary and backup hubs subject to the installed hubs, allocation of defensive resources, and the risk of disruptions. The formulation with the location of hubs, allocation of protection budget, flow routing between two nodes in origin-destination via the primary and backup hubs, and hub failure probabilities have not been remarked in the literature. As the problem is an NP-hard, the performance of some metaheuristic algorithms called tabu search (TS), simulated annealing (SA), variable neighborhood search (VNS), imperialist competitive algorithm (ICA) and genetic algorithm (GA) is investigated to solve instances of problem with 50 nodes and 5, 10, and 15 hubs. Computational results show that the SA and TS algorithms are superior to existing metaheuristic methods to the novel problem based on solution accuracy and computational time, respectively. Additionally, this study presents a fast and robust hybrid approach that combines the advantages of TS and SA. The proposed algorithm has been successfully used to solve a large number of instances of this problem via sensitivity analysis and instances from the Turkish Postal data set. Several experimental results indicate the applicability of the new model and the advantage of the new hybrid method compared to other metaheuristic algorithms concerning decision and execution time. (C) 2021 Elsevier B.V. All rights reserved.
Vafashoar, RezaMorshedlou, HosseinMeybodi, Mohammad Reza
21页
查看更多>>摘要：The flying speed and trajectory of particles are subject to several factors, including neighborhood structures, inertia weight, and acceleration coefficients. This paper improves particle swarm optimizer by exploiting these factors. Specifically, it proposes an approach to adjust the neighborhood structures of particles adaptively. The search task is divided between two groups of particles, termed even and uneven, to perform a vigorous in-depth search. Each particle group pursues a different objective and conducts its search in a different manner. Even particles adaptively adjust their number of attractors and neighborhood radiuses to experience various flying trajectories and paces. Each uneven particle follows a single even one for a while until it is assigned to another even particle. Uneven particles are responsible for performing fine-grained searches in the vicinity of their associated even particles as well as their previously experienced locations. A tree structure is utilized to implement the neighborhood structures of the proposed method. In the presented structure, particles can experience large neighborhoods by choosing their attractors from higher levels of the tree. The proposed method is experimentally investigated on the comprehensive CEC2013 benchmark set and two challenging real-world problems: non-uniform circular antenna array synthesis and image segmentation. The comparison results with advanced particle swarm optimization algorithms demonstrate that search bifurcation and topology adjustment can significantly improve particle swarm optimization. Experimental results also indicate that the proposed method can be successfully employed for solving challenging real-world problems with various characteristics. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：The recognition of epithelial type 2 (HEp-2) cells has become an important tool for detecting autoimmune diseases diagnosis by using indirect immunofluorescence (IIF) images. For many medical image tasks, the accurate segmentation is considered as the primary step for the classification task, which inspires researchers to solve problems by jointly implementing multiple tasks. For this reason, we devise a hybrid network architecture, which combines the segmentation and classification networks for the final classification of HEp-2 cells, where a multi-task generative adversarial networks (GANs) is employed to produce accurate segmentation masks for improving the latter classification performance. Specifically, the devised generation and discrimination subnetworks establish the GANs architecture for the accurate segmentation masks of HEp-2 cells. Also, the original images are utilized as the conditional input, which are concatenated by the generated masks and the ground-truth to train the discriminator for two tasks: the first task determines whether the generated mask is a ground-truth or not and the other one distinguishes the category of the processed HEp-2 cell image. Furthermore, the ResNet-34 and MobileNetv3 are used as segmentation and classification base network, respectively. We modify the MobileNetv3 structure by adding the channel of the middle outputs, which is called augmented channel MobileNetv3 (ACM-Net). Both the discriminator and classifier share the weights of ACM-Net. The extensive experiments on the public ICPR 2016 task1 dataset show that the proposed hybrid-task based GANs structure can obtain promising segmentation and classification performance via jointly training mode, which achieves a Dice of 97.04% and for segmentation and a prediction accuracy of 98.82% for classification, respectively. (C) 2021 Published by Elsevier B.V.
NSTL
Elsevier