Lin, ZhipengGao, ZhiJi, HongZhai, Ruifang...
13页查看更多>>摘要：Automatic classification of cervical cells plays a critical role in the Computer-assisted Cytology Test (CCT) system. The efficiency of the CCT system can be promoted by sacrificing the microscopic image resolution to speed up the microscopic image acquisition. In this case, the low resolution of the cell image will severely deteriorate the performance of available Convolutional Neural Networks (CNN) based classification methods. Inspired by the positive effect of super-resolution in addressing classification or recognition tasks, we propose a cervical cell classification algorithm leveraging simultaneous superresolution, which is achieved using Generative Adversarial Network (GAN) techniques. Our framework is designed in an end-to-end manner wherein the classification loss is back-propagated into the superresolution network during training. Moreover, we perform ordinal regression with smooth L1 loss to further improve the classification results. Extensive experiments have verified the effectiveness of our method. Our simultaneous super-resolution based method achieves 93.5% classification accuracy on the 6-class Heer dataset, outperforming the method using only the state-of-the-art classifier by an obvious margin of 3.2%. Besides, our ordinal regression method significantly improves the MAE (Mean Absolute Error) by 0.0143 and 1-off accuracy by 0.95% on the 4-class Heer dataset. For the Herlev dataset, our method yields the classification accuracy of 98.1% and 97.6% for the 2-class and 7-class problems, which is still competitive even with low-resolution input. (c) 2021 Elsevier B.V. All rights reserved.
原文链接:
NSTL
Elsevier
Herrera, LeonardoRodriguez-Linan, M. C.Meza-Sanchez, MarlenMonay-Arredondo, Luis...
13页查看更多>>摘要：The present work focuses on an open problem in the design of Extended Kalman filters: the lack of knowledge of the measurement noise covariance. A novel extension of the analytic behaviors framework, which integrates a theoretical formulation and evolutionary computing, has been introduced as a design methodology for the construction of this unknown parameter. The proposed methodology is developed and applied for the design of Evolved Extended Kalman Filters for nonlinear first-order dynamical systems. The proposed methodology applies an offline evolutionary synthesis of analytic nonlinear functions, to be used as measurement noise covariance, aiming to minimize the Kalman criterion. The virtues of the methodology are exemplified through a complex, highly nonlinear, firstorder dynamical system, for which 2649 optimized replacements of the measurement noise covariance are found. Under different scenarios, the performance of the Evolved Extended Kalman Filter with unknown measurement noise covariance is compared with that of the conventional Extended Kalman Filter where the measurement noise covariance is known. The robustness of the Evolved Extended Kalman Filter is demonstrated through numerical evaluation.
原文链接:- NSTL
- Elsevier
Zhu, XishunLai, ZhengliangLiang, YaruXiong, Jianping...
17页查看更多>>摘要：Image hiding is the process of hiding a secret image in another meaningful image or other carriers so that the secret image remains imperceptible and can be recovered securely at the receiving end. The output image of an image hiding algorithm hides the secret image and visually appears to be the same as the carrier image, thus reducing the possibility of being attacked. The current hiding algorithms have relatively low hiding capacity and weak security. In this paper, we propose a generative image hiding algorithm based on a residual convolutional neural network (ResCNN) in wavelet domain to overcome the above-mentioned shortcomings. First, the secret image was subjected to wavelet transform. The low-frequency band of wavelet coefficients were discarded, and only the high-frequency bands were retained as features. These features were then effectively embedded into the carrier image by a generative ResCNN. The recovery network was trained simultaneously with the hiding network so as to extract the hidden features from the container and reconstruct the secret image. Pixel shuffle was used to recover a high-resolution secret image. The experimental results show that the proposed image hiding algorithm is capable of obtaining state-of-the-art results in terms of high hiding capacity and strong security measures. (C) 2021 Elsevier B.V. All rights reserved.
原文链接:- NSTL
- Elsevier
Chen, Ze-huiWan, Shu-pingDong, Jiu-ying
19页查看更多>>摘要：Since makeshift hospitals have strong ability in blocking the spread of the virus, how to design some methods to select the reasonable sites of makeshift hospitals is vitally important for containing COVID-19. This paper investigates an efficiency-based multi-criteria group decision making (MCGDM) method by combining the best-worst method (BWM) and data envelopment analysis (DEA) in trapezoidal interval type-2 fuzzy (TrIT2F) environment. This MCGDM method is called TrIT2F-BWM-DEA, where the TrIT2F-BWM is used to determine the weights of criteria and decision-makers, and the TrIT2F-DEA is employed to rank alternatives by measuring their overall efficiencies. Based on cut set theory, the expectation and average expectation (AE) of TrIT2FSs are successively defined. To solve three key issues in the development of the TrIT2F-BWM, this paper proposes a flexible ranking relation of TrIT2FSs to transform the TrIT2F constraints, initiates an efficient theorem to normalize the TrIT2F weights, and designs an input-based consistency ratio to check the reliability of the determined weights. A fully TrIT2F-DEA model is originally built to measure the TrIT2F efficiencies of alternatives. The alternatives are finally ranked according to the AEs of alternatives' TrIT2F efficiencies. A site selection case of Fangcang hospitals and some comparative analyses are provided to confirm the validity and merits of the proposed TrIT2F-BWM-DEA. (C) 2021 Elsevier B.V. All rights reserved.
原文链接:- NSTL
- Elsevier
Han, YuxinYan, XueliGu, Xingsheng
16页查看更多>>摘要：The multi-stage multi-product batch plant scheduling problem is an important part of batch chemical industry scheduling problems. Different from the multi-purpose batch scheduling problem, this problem can be characterized by multiple stages with non-identical parallel units, and multiple batches of customer orders. Numerous methodologies for this problem have been investigated to addressing scheduling cases of different production systems in the past two decades. This paper focus on the large-scale batch plant scheduling problem by minimizing the make-span in the scheduling horizon. And a novel hybrid discrete differential evolutionary algorithm is proposed to handle this problem. First, a novel two-line encoding scheme is constructed based on discrete and continuous variables. The sequence of orders is represented by a time-based representation method. Second, two novel mutation methods are proposed within the framework of encoding method. Two methods provide multiple search directions which helps improve the exploration ability and diversity of the population. At last, two local permutation methods are applied to improve the local optimal for the algorithm. The proposed work is tested through several real industrial instances with different sizes and characteristics by comparing with mixed integer linear programming method and meta-heuristic algorithms. The improvement of proposed work is also analyzed by the instances. The results report the efficiency and effectiveness of the novel proposed evolutionary algorithm in solving large scale multi-stage multi-product batch plant scheduling problem.
原文链接:- NSTL
- Elsevier
Zhang, SitongLi, YibingDong, Qianhui
12页查看更多>>摘要：Path planning is one of the most essential part in autonomous navigation. Most existing works suppose that the environment is static and fixed. However, path planning is widely used in random and dynamic environment (such as search and rescue, surveillance and other scenarios). In this paper, we propose a Deep Reinforcement Learning (DRL)-based method that enables unmanned aerial vehicles (UAVs) to execute navigation tasks in multi-obstacle environments with randomness and dynamics. The method is based on the Twin Delayed Deep Deterministic Policy Gradients (TD3) algorithm. In order to predict the impact of the environment on UAV, the change of environment observations is added into the Actor-Critic network input, and the two-stream Actor-Critic network structure is proposed to extract features of environment observations. Simulations are carried out to evaluate the performance of the algorithm and experiment results show that our method can enable the UAV to complete autonomous navigation tasks safely in multi-obstacle environments, which reflects the efficiency of our method. Moreover, compared to DDPG and the conventional TD3, our method has better generalization ability. (C) 2021 Elsevier B.V. All rights reserved.
原文链接:- NSTL
- Elsevier
Arulprakasam, SakthidasanMuthusamy, Senthilkumar
9页查看更多>>摘要：The sharp escalating power demand and configuration changes in distribution networks (DNs) may operate the networks more closely to voltage stability boundaries. Under critical operating conditions, the DN is not able to provide good voltage profile and may experience voltage collapse. The performances of the DN can be improved by optimally reconfiguring the network. This paper models the reconfiguration of DN as an optimization problem with objectives of lowering active power loss, improving the voltage profile and enhancing the voltage stability; and suggests a new reconfiguration method involving rain-fall optimization and non-dominated sorting to obtain the best compromised solution for DNs. It presents simulation results of standard 33-, 69- and 95-node DNs, and exhibits that the method was able to lower the active power loss from 201.97 kW to 139.5525 kW, from 225 kW to 98.6082 kW and from 89.6733 kW to 30.5700 kW for 33, 69 and 95 node DN systems respectively. In a similar way, it portrays that the method was able to produce better results in enhancing the voltage profile and voltage stability.
原文链接:- NSTL
- Elsevier
Das, AbhishekMohapatra, Saumendra KumarMohanty, Mihir Narayan
11页查看更多>>摘要：Research on biomedical science has many components like biomedical engineering, biomedical signal processing, gene analysis, and biomedical image processing. Classification, detection, and recognition have a great value for disease diagnosis and analysis. In this work, biomedical image classification is discussed. In one part, the brain tumor is considered with brain magnetic resonance images and in the other part, COVID affected chest X-rays have been classified using the ensemble approach. The images have been collected from Kaggle online platform. For this purpose, four heterogeneous base classifiers as Convolutional Neural Network, Recurrent Neural Network, Long Short Term Memory, and Gated Recurrent Unit are considered, and metadata is generated. Further, for the detection purpose, a fuzzy min-max model is utilized to avoid uncertainty. The ensemble output from the base classifiers is fed to the fuzzy model in terms of class probability and labels. The min-max algorithm for correct decisions is used in the fuzzy model. The measuring parameters like precision, recall, accuracy, sensitivity, specificity, and F1-score are evaluated. 100% training accuracy for both the datasets is obtained whereas 97.62% and 95.24% of validation accuracy are found for brain image and chest X-ray image classification respectively as exhibited in the result section. (C) 2021 Elsevier B.V. All rights reserved.
原文链接:- NSTL
- Elsevier
Sahargahi, VahidehMajidnezhad, VahidAfshord, Saeid TaghaviJafari, Yasser...
36页查看更多>>摘要：Increasing complexity of real world problems motivated an area to explore efficient optimization methods to solve such problems. Existing optimization algorithms cannot solve all type of problems efficiently (NFL theorem), so new algorithms are proposed to find the better solutions for such complex optimization problems. However, their efficiency and performance can be still improved. Therefore, to follow this vital purpose, in this paper, a novel metaheuristic algorithm, called intelligent clonal optimizer (ICO), is proposed to solve continuous optimization problems. In the proposed algorithm, the initial population is generated through the chaos theory to enhance its exploration capability. It lacks any crossover operator. Instead, a novel clonal operator copying candidate solutions according to their fitness in a self-adaptive way is proposed. Cloning each parent is carried out by two methods, and according to these methods, each offspring is located near the parent or in direction of temporary target. The offsprings are classified to two classes. In addition, a novel conservative selection operator is proposed. According to this operator, the new population is selected from two classes of offsprings and current population by maintaining population diversity. The performance of the ICO algorithm is assessed on 39 well-known unimodal, multimodal, fixed-dimensional multimodal, composite and CEC2019 benchmark functions as well as three engineering application problems. Results of the proposed ICO are compared to sixteen state-of-art metaheuristic algorithms in three categories including the most well-known and recently developed algorithms and the best performer of IEEE CEC competitions using statistical analysis, scalability analysis, Wilcoxon Signed-Rank Test, Friedman test, computational time analysis and convergence analysis. The obtained results proved that ICO performs better than state-of-art metaheuristics in sense of scalability and accurate convergence. According to average rank of Friedman test, the proposed ICO is firstly ranked among others. (c) 2021 Elsevier B.V. All rights reserved.
原文链接:- NSTL
- Elsevier
