首页期刊导航|Applied Soft Computing
期刊信息/Journal information
Applied Soft Computing
Elsevier Science, B.V.
Applied Soft Computing

Elsevier Science, B.V.

1568-4946

Applied Soft Computing/Journal Applied Soft ComputingEIISTPSCIAHCI
正式出版
收录年代

    An attention based dual learning approach for video captioning

    Ji, WantingWang, RuiliTian, YanWang, Xun...
    9页
    查看更多>>摘要:Video captioning aims to generate sentences/captions to describe video contents. It is one of the key tasks in the field of multimedia processing. However, most of the current video captioning approaches utilize only the visual information of a video to generate captions. Recently, a new encoder-decoderreconstructor architecture was developed for video captioning, which can capture the information in both raw videos and the generated captions through dual learning. Based on this architecture, this paper proposes a novel attention based dual learning approach (ADL) for video captioning. Specifically, ADL is composed of a caption generation module and a video reconstruction module. The caption generation module builds a translatable mapping between raw video frames and the generated video captions, i.e., using the visual features extracted from videos by an Inception-V4 network to produce video captions. Then the video reconstruction module reproduces raw video frames using the generated video captions, i.e., using the hidden states of the decoder in the caption generation module to reproduce/synthesize raw visual features. A multi-head attention mechanism is adopted to help the two modules focus on the most effective information in videos and captions, and a dual learning mechanism is adopted to fine-tune the performance of the two modules to generate final video captions. Therefore, ADL can minimize the semantic gap between raw videos and the generated captions by minimizing the differences between the reproduced and the raw videos, thereby improving the quality of the generated video captions. Experimental results demonstrate that ADL is superior to the state-of-the-art video captioning approaches on benchmark datasets. (C) 2021 Published by Elsevier B.V.

    Distributed multi-objective grey wolf optimizer for distributed multi-objective economic dispatch of multi-area interconnected power systems

    Yin, LinfeiSun, Zhixiang
    15页
    查看更多>>摘要:With the gradual opening and rapid development of power markets, large-scale multi-area interconnected power systems (LMIPSs) have become an inevitable pattern. The traditional centralized economic dispatch optimization method has the disadvantages of slow calculation speed, easy exposure of private equipment information, and considers only one cost objective. This paper introduces the distributed concept into the multi-objective grey wolf optimizer (MOGWO) to mitigate these deficiencies; then proposes the distributed MOGWO (DMOGWO). When the DMOGWO solves the LMIPS problems, the sub-problems of each area are optimized independently, and the overall optimization can be realized by sharing only part of the boundary bus information between areas. Case studies are carried out in two cases of the Institute of Electrical and Electronics Engineers (IEEE) 39-bus and 118-bus systems. The results show that when solving the multi-objective economic dispatch in LMIPS, compared with centralized optimization, the proposed DMOGWO can effectively ensure the privacy of information, the obtained objective values are smaller, and the performance test is better.(C)& nbsp;2021 Elsevier B.V. All rights reserved.

    Society-based Grey Wolf Optimizer for large scale Combined Heat and Power Economic Dispatch problem considering power losses

    Hosseini-Hemati, SamanBeigvand, Soheil DerafshiAbdi, HamdiRastgou, Abdollah...
    15页
    查看更多>>摘要:Combined Heat and Power Economic Dispatch (CHPED) reflects a momentous optimization and operation problem in the power systems for optimally allocating the produced heat and power to the committed Power-only (PO), Heat-only (HO), and CHP units. In this work, a new societybased optimization algorithm using social hierarchy of grey wolves, namely Society-based Grey Wolf Optimizer (SGWO) is proposed for the CHPED problem. This scheme divides the group of wolves to some societies so that each society has its own leader. These leaders follow the dominant wolf and also guide their societies in which the classification of wolves is based on the social hierarchy. Moreover a novel attacking and hunting the prey is suggested to change the effects of dominant, ordinate, and subordinate wolves in their new roles in societies. In order to verify the capabilities of the SGWO, simulations are conducted through two large-scale CHPED problems with the related challenges like Valve-Point Loading Effect (VPLE) and Prohibited Zones (PZ) of PO units, mutual dependency of produced heat and power of CHP units, and especially transmission power losses. Moreover, it is evaluated on twenty-three standard functions to verify its stability on different low- and highdimensional functions. Comparisons based on the obtained solutions by different methods demonstrate the robustness and superior performance of the presented technique to fast provide better optimum point (more economical benefits) and solution quality meeting all equality and inequality constraints. In addition, the results of CHPED indicate that approximate to 0.5% reduction in the production cost results in up to $2.6 - $34 x 10(6) increase in Annual Cost Saving (ACS). Also, the transmission losses and PZs of POs can increase the production cost by about 1.4% which leads to over $1.3 x 10(7) reduction in ACS in comparison with ignoring them. Moreover, this condition increases the computational time by about 35% while the proposed method can still be up to 13- 26 times faster than the other analysed algorithms. (C) 2021 Elsevier B.V. All rights reserved.

    A least square support vector machine approach based on bvRNA-GA for modeling photovoltaic systems

    Liu, XiuWang, NingMolina, DanielHerrera, Francisco...
    21页
    查看更多>>摘要:Accurate model plays an important role in designing, assessing, and controlling photovoltaic (PV) systems. In this work, the least-squares support vector machine (LSSVM) is adopted to model the current-voltage (V-I) characteristic curves of different PV systems. A novel RNA genetic algorithm (bvRNA-GA) is proposed to determine the parameters of LSSVM. The bvRNA-GA is featured by designing the bulge loop crossover operator and the virus-induced mutation operator, they are employed to balance the exploration and exploitation capacities. Different experiments with 10 benchmark functions are conducted to show that the search efficiency of bvRNA-GA is better than the other four state-of-art algorithms. The outputs of bvRNA-GA optimized LSSVM models can better agree with the real outputs of different PV systems, the modeling results demonstrate the effectiveness of bvRNA-GA in solving real-world problems. (C) 2021 Elsevier B.V. All rights reserved.

    Fuzzy temporal convolutional neural networks in P300-based Brain-computer interface for smart home interaction

    Verga, Christian FloresQuevedo, JonathanEscandon, ElmerKiani, Mehrin...
    11页
    查看更多>>摘要:The processing and classification of electroencephalographic signals (EEG) are increasingly performed using deep learning frameworks, such as convolutional neural networks (CNNs), to generate abstract features from brain data, automatically paving the way for remarkable classification prowess. However, EEG patterns exhibit high variability across time and uncertainty due to noise. It is a significant problem to be addressed in P300-based Brain Computer Interface (BCI) for smart home interaction. It operates in a non-optimal natural environment where added noise is often present. In this work, we propose a sequential unification of temporal convolutional networks (TCNs) modified to EEG signals, LSTM cells, with a fuzzy neural block (FNB), we called EEG-TCFNet. Fuzzy components may enable a higher tolerance to noisy conditions. We applied three different architectures comparing the effect of using block FNB to classify a P300 wave to build a BCI for smart home interaction with healthy and post-stroke individuals. Our results reported a maximum classification accuracy of 98.6% and 74.3% using the proposed method of EEG-TCFNet in subject-dependent strategy and subject-independent strategy, respectively. Overall, FNB usage in all three CNN topologies outperformed those without FNB. In addition, we compared the addition of FNB to other state-of-the-art methods and obtained higher classification accuracies on account of the integration with FNB. The remarkable performance of the proposed model, EEG-TCFNet, and the general integration of fuzzy units to other classifiers would pave the way for enhanced P300-based BCIs for smart home interaction within natural settings. (C) 2021 Elsevier B.V. All rights reserved.

    OPFaceNet: OPtimized Face Recognition Network for noise and occlusion affected face images using Hyperparameters tuned Convolutional Neural Network

    Lokku, GurukumarReddy, G. HarinathaPrasad, M. N. Giri
    21页
    查看更多>>摘要:Face recognition is considered as important research in computer vision applications, and it is regarded as the basic biometric security system. Research related to face recognition has been done in the past several years. Still, many more challenges associated with this field need to be addressed. Some literary works have designed the face recognition model on the relatively controlled environments; yet, their performance in general settings has been substandard This paper develops an Optimal Face Recognition Network (OPFaceNet) to recognize the face images affected by high noise and occlusion. The feature patterns subjected to noise like LBP, FLBP, and NRLBP are extracted. The average of all three patterns is given to the proposed Convolutional Neural Network (CNN) classifier. As the main contribution, the CNN model is enhanced by optimizing the Fitness Sorted Rider Optimization Algorithm (FS-ROA). This algorithm optimizes the hyperparameters of CNN like Convolutional layer, Pooling Layer, Fully connected layer, number of Hidden layers, and Type of Pooling. Finally, the simulation results show that the system achieves a good recognition rate of 97.2% and is robust against variations in terms of occlusion and noise when benchmarked over diverse datasets. (C) 2021 Elsevier B.V. All rights reserved.

    An adaptive artificial bee colony with reinforcement learning for distributed three-stage assembly scheduling with maintenance

    Wang, JingLei, DemingCai, Jingcao
    13页
    查看更多>>摘要:Distributed three-stage assembly scheduling problem extensively exists in the real-life assembly production process and is seldom considered. The integration of reinforcement learning with metaheuristic can effectively improve the performance of meta-heuristic and effectively solve the problem; however, the integration is seldom used to cope with the problem. In this study, distributed three-stage assembly scheduling problem with DPm -> 1 layout and maintenance at three stages is considered and a mathematical model is provided. A new artificial bee colony with Q-learning (QABC) is proposed to minimize maximum tardiness. An effective Q-learning algorithm is implemented to dynamically select search operator, which consists of 12 states based on population quality evaluation, 8 actions defined by global search and neighborhood search, a new reward and an effective action selection. Two employed bee swarms are formed, an adaptive communication and an adaptive competition process between them are adopted to intensify exploration ability and improve search efficiency. QABC and its four comparative algorithms are tested on 80 instances. The computational results demonstrate that the new strategies of QABC really improve its search performance and QABC is a competitive algorithm for the considered problem. (C) 2021 Published by Elsevier B.V.

    Social network clustering and consensus-based distrust behaviors management for large-scale group decision-making with incomplete hesitant fuzzy preference relations

    Lu, YanlingXu, YejunHuang, JingWei, Ju...
    17页
    查看更多>>摘要:With the development of social network platforms, large-scale group decision-making in social network (LSGDM-SN) has been formed. As decision makers (DMs) come from different fields and have complex individual backgrounds, which leads to their distrust in the moderator. Moreover, in LSGDMSN, since DMs can hardly grasp all the information about the decision problem, the hesitant fuzzy preference relations (HFPRs) they have expressed may be incomplete. However, in current LSGDM-SN issues, the distrust behaviors and incomplete HFPRs have never been discussed simultaneously. In this context, this paper aims to propose a method to estimate incomplete values in HFPRs, and develop a consensus management process which considers distrust behaviors. This paper focuses on LSGDM-SN on the basis of social network clustering and consensus-based distrust behaviors management with incomplete HFPRs. In this paper, a social network clustering method based on grey clustering algorithm is proposed to classify the DMs with similar social clustering degree into a subset. Afterwards, a method including two situations is developed to estimate incomplete values in HFPRs. Furthermore, an identification mechanism is presented to detect the DMs' distrust behaviors, and three modification strategies are provided for managing different types of distrust behaviors. In addition, a case study is given to illustrate the feasibility of the proposed method. Finally, comparative analysis and discussion are explored to verify the advantages of the proposed LSGDM-SN with incomplete HFPRs. (C) 2021 Elsevier B.V. All rights reserved.

    A grid-guided particle swarm optimizer for multimodal multi-objective problems

    Qu, BoyangLi, GuosenYan, LiLiang, Jing...
    26页
    查看更多>>摘要:This paper proposes a grid-guided particle swarm optimizer for solving multimodal multi-objective optimization problems that may have multiple disjoint Pareto sets corresponding to the same Pareto front. The concept of grid in the decision space is adopted to detect the special promising subregions, and accordingly to generate multiple subpopulations. The grid-guided technique can maintain the diversity of the population during the search process and improve the search efficiency. To obtain a well distributed Pareto optimal set, an external archive maintenance strategy is employed to select and store the solutions found in each generation. In addition, nine new multimodal multi-objective benchmark test functions are designed. The proposed algorithm is compared with ten state-of-the-art evolutionary algorithms on thirty-seven test functions. Moreover, the proposed algorithm is applied to solve a real-world problem. The experimental results demonstrate that the proposed algorithm is able to achieve superior performance compared with the alternative evolutionary methods considered. (C) 2021 Elsevier B.V. All rights reserved.

    Detecting adversarial examples by positive and negative representations

    Luo, WenjianWu, ChenwangNi, LiZhou, Nan...
    13页
    查看更多>>摘要:Deep neural networks (DNNs) have been successfully applied in various fields. However, it has been demonstrated that a well-designed and quasi-imperceptible perturbation can confuse the targeted DNNs classifier with high confidence and lead to misclassification. Examples with such perturbations are called adversarial examples, and it is a challenging task to detect them. In this paper, we propose a positive-negative detector (PNDetector) to detect adversarial examples. The PNDetector is based on a positive-negative classifier (PNClassifier), which is trained by both the original examples (called positive representations) and their negative representations with the same structural and semantic features. The principle of the PNDetector is that the feature space of the positive and negative representations of adversarial examples under the PNClassifier has a high probability of belonging to different categories, while its performance on clean examples is not reduced by adding negative example representations into the train set. We test the PNDetector with adversarial examples generated by eight typical attack methods on four typical datasets. The experimental results demonstrate that the proposed detector is efficient in all datasets and under all attack types. Furthermore, its detection performance is comparable to that of state-of-the-art methods. (C) 2021 Elsevier B.V. All rights reserved.