查看更多>>摘要：Early detection of Tuberculosis or TB can help in mitigating the chances of affecting the other body parts like the kidney, spine and brain, thereby reducing the death rate due to this disease. However, manual diagnosis by radiologists using Chest X-rays may include human error. Therefore, researchers have been trying hard to develop a computerized decision support system for efficient detection of TB from Chest X-ray images. In this work, we have proposed a model for screening TB using Chest X-ray images where the decisions from three base learners are combined using the type-1 Sugeno fuzzy integral based ensemble technique. Fuzzy measures required in this fuzzy integral based ensemble method are set experimentally in many state-of-the-art works. To overcome such manual tuning, we have used meta-heuristic optimization algorithms to set the fuzzy measures optimally during the training process of the model. The performance of the ensemble technique on the validation set is considered as the decider of the optimal fuzzy measures. Before applying the ensemble method we have extracted features from images using three state-of-the-art deep learning models, namely DenseNet121, VGG19 and ResNet50 pre-trained on imageNet dataset. With the above pre-trained models, the base learners are built using additional fully connected and softmax layers. We have evaluated the present work on a new and publicly available TB dataset consisting of Chest X-ray images. The obtained results (irrespective of the optimizer used) confirm that our method has outperformed state-of-the-art methods used for TB classification. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：The data-driven nature of deep learning (DL) models for semantic segmentation requires a large number of pixel-level annotations. However, large-scale and fully labeled medical datasets are often unavailable for practical tasks. Recently, partially supervised methods have been proposed to utilize images with incomplete labels in the medical domain. To bridge the methodological gaps in partially supervised learning (PSL) under data scarcity, we propose Vicinal Labels Under Uncertainty (VLUU), a simple yet efficient framework utilizing the human structure similarity for partially supervised medical image segmentation. Motivated by multi-task learning and vicinal risk minimization, VLUU transforms the partially supervised problem into a fully supervised problem by generating vicinal labels. We systematically evaluate VLUU under the challenges of small-scale data, dataset shift, and class imbalance on two commonly used segmentation datasets for the tasks of chest organ segmentation and optic disc-and-cup segmentation. The experimental results show that VLUU can consistently outperform previous partially supervised models in these settings. Our research suggests a new research direction in label-efficient deep learning with partial supervision. (C) 2021 The Authors. Published by Elsevier B.V.
查看更多>>摘要：The novel Coronavirus disease 2019 (COVID-2019) has become a global pandemic and affected almost all aspects of our daily life. The total number of positive COVID-2019 cases has exponentially increased in the last few months due to the easy transmissibility of the virus. It can be detected using the nucleic acid test or the antibodies blood test which are not always available and take several hours to get the results. Therefore, researchers proposed computer-aided diagnosis systems using the state-of-theart artificial intelligence techniques to learn imaging biomarkers from chest computed tomography and X-ray radiographs to effectively diagnose COVID-19. However, previous methods either adopted transfer learning from a pre-trained model on natural images or were trained on limited datasets. Either cases may lead to accuracy deficiency or overfitting. In addition, feature space suffers from noise and outliers when collecting X-ray images from multiple datasets. In this paper, we overcome the previous limitations by firstly collecting a large-scale X-ray dataset from multiple resources. Our dataset includes 11,312 images collected from 10 different data repositories. To alleviate the effect of the noise, we suppress it in the feature space of our new dataset. Secondly, we introduce a supervision mechanism and combine it with the VGG-16 network to consider the differences between the COVID-19 and healthy cases in the feature space. Thirdly, we propose a multi-site (center) COVID19 graph convolutional network (GCN) that exploits dataset information, the status of training samples, and initial scores to effectively classify the disease status. Extensive experiments using different convolutional neural network-based methods with and without the supervision mechanism and different classifiers are performed. Results demonstrate the effectiveness of the proposed supervision mechanism in all models and superior performance with the proposed GCN. (c) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Highly accurate forecasting of particulate matter concentration (PMC) is essential and effective for establishing a reliable air pollution early warning system and has both theoretical and practical significance. To meet this demand, a novel multi-scale hybrid learning framework based on robust local mean decomposition (RLMD) and moving window (MW) ensemble strategy is developed for PM2.5 and PM10 forecasting. In this architecture, the RLMD is adopted to adaptively decompose the PMC time series (PMCTS) into several production functions and one residue with different frequencies. These subseries are simpler than the original PMCTS, but they still work alongside mode aliasing. Thus, following the well-established "linear and nonlinear" modeling philosophy, a novel hybrid learning framework, composed of the autoregressive integrated moving average (ARIMA) and combined kernel function relevance vector machine (RVMcom), is proposed to capture both the linear and nonlinear patterns in the subseries. To obtain better final outputs, based on the definition of the ensemble improvement degree, the MW ensemble method is used to merge the forecasting results of all subseries. A comprehensive experiment is conducted using PM2.5 and PM10 datasets from four municipalities in China to investigate the forecasting performance of our proposed framework, and the results demonstrate that our proposed RLMD-ARIMA-RVMcom-MW (R-A&R-com-M) model is superior to other considered methods in terms of forecasting accuracy and generalization ability. This means that the developed forecasting architecture has a great application value in the field of PMCTS prediction. (C) 2021 Published by Elsevier B.V.
查看更多>>摘要：Visual detection is an available approach for measuring gear pitting. Unfortunately, the number of gear pitting images is limited, resulting in that the detection accuracy of gear pitting is unsatisfactory. In order to augment gear pitting samples with different styles, a novel Cycle Generative Adversarial Network based on a symmetric tree structure (Tree-CycleGAN) is proposed. In Tree-CycleGAN, a new type of generator with tree structure named tree generator is designed to produce various types of high quality target samples from the source-domain samples, and a maximum diversity loss is constructed to enlarge the difference between two arbitrary branches; then a similar tree reconstructor is designed for translating target samples into source samples. Two discriminators are designed for making the generated images approximate to the target images in two cyclic processes. Via inception score, structural similarity indexes and peak-signal-to-noise ratio, the quality and diversity of images obtained by Tree-CycleGAN are evaluated. Comparative results show the superiority of Tree-CycleGAN over other domain adaptation GANs. The proposed Tree-CycleGAN combined with U-net has been successfully applied to gear pitting detection. Experimental results prove that the proposed methodology precedes the basic U-net method without sample augmentation and the method based on CycleGAN and U-net. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：This paper proposes two hierarchical evolutionary optimization methods based on variable fidelity analysis and search space contraction for aerodynamic shape design, i.e. hierarchical evolutionary Pareto and Nash games. One of these techniques is used in the optimization process, namely the advantages of high and low fidelity flow simulation. The high-fidelity model provides solution accuracy while the low-fidelity model reduces the computational cost. Especially, the search space contraction and the population size reduction are introduced in the process of transition from the optimization on low fidelity simulation to the optimization on high fidelity simulation, so that the optimization based on high fidelity simulation can get the high-precision optimal solution with a relatively less Central Processing Unit(CPU) cost. They are applied to the single objective natural laminar wing shape design at transonic flow and the multidisciplinary shape optimization of a hypersonic air-breathing vehicle respectively. The optimization results show that regardless of a single objective or multiobjective/multidisciplinary optimization problem, the new hierarchical optimization methods proposed in this paper can improve the optimization efficiency by 5-10 times. (C) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：Most of the practical engineering optimization problems are highly nonlinear, nonconvex, and sometimes discontinuous. Classical optimization techniques, mostly being differential calculus based, either fail to find the optimal solution for practical problems or provide solution after relaxing the nonlinearities. Over the time, population-based meta-heuristic techniques have gained enough popularity among research fraternity due to their unrestricted performance on the nature of the optimization problems. Despite their better performances, they sometimes suffer from the problem of trapping into local optima. Hence, suitable strategies should be adopted to overcome the above said issues. In view of this, a new Levy Flight (LF) based Adaptive Particle Swarm Optimization (APSOLF) technique is designed and proposed in this work to solve complex, nonlinear optimization problems. The performance of the proposed algorithm is gauged by applying it on various mathematical benchmark functions. The technique is also applied to solve the practical electrical engineering problems where the task of proposed algorithm is to optimize the Static Synchronous Series Compensator (SSSC) parameters to improve the Maximum Loadability Limit (MLL) of some standard test power systems viz. IEEE 14, 30, 57, 118 and a practical Indian southern region 205 buses. The results obtained are compared with other variants of PSO. Furthermore, the robustness of the proposed algorithm is tested by performing the statistical analysis (both parametric and nonparametric tests). The results confirm better efficiency, robustness, and consistency of the proposed algorithm. The simulations are performed on MATLAB environment. (C) 2021 Published by Elsevier B.V.
查看更多>>摘要：Droughts are the most spatially complex geohazard, which often lasts for years, thereby severely impacting socio-economic sectors. One of the critical aspects of drought studies is developing a reliable and robust forecasting model, which could immensely help drought management planners in adopting adequate measures. Further, the prediction of drought events are extremely challenging due to the involvement of several hydro-meteorological factors, which are further aggravated by the effect of climate change. Among the several techniques such as statistical, physical and data-driven that are used to forecast droughts, artificial neural networks provide one of the most robust approach. As droughts are inherently non-linear and multivariate in nature, the capability of neural networks to capture the dynamic relationship easily and efficiently has seen a rise in its use. Here we evaluate the most used architectures in the last two decades, using scientometric analysis. A general framework used in drought prediction studies is explained and examples from various continents are provided, thus exploring the topic in a global context. The findings show that using sophisticated input representation, the artificial intelligence-based solutions applied to drought prediction of hydro-meteorological variables have promising success, particularly in complex geographical scenarios. The future works need to focus on interpretable models, use of deep learning architectures for long lead time forecasting and use of neural networks to predict different drought characteristics like drought propagation and flash droughts. We also summarize the most widely used neural network approaches in spatial drought prediction, which would serve as a foundation for future research in drought prediction studies. (C) 2021 Elsevier B.V. All rights reserved.
Abba, S., IAbdulkadir, R. A.Sammen, Saad ShPham, Quoc Bao...
17页
查看更多>>摘要：The establishment of water quality prediction models is vital for aquatic ecosystems analysis. The traditional methods of water quality index (WQI) analysis are time-consuming and associated with a high degree of errors. These days, the application of artificial intelligence (AI) based models are trending for capturing nonlinear and complex processes. Therefore, the present study was conducted to predict the WQI in the Kinta River, Malaysia by employing the hybrid AI model i.e., GA-EANN (genetic algorithm-emotional artificial neural network). The extreme gradient boosting (XGB) and neuro-sensitivity analysis (NSA) approaches were utilized for feature extraction, and six different model combinations were derived to examine the relationship among the WQI with water quality (WQ) variables. The efficacy of the proposed hybrid GA-EANN model was evaluated against the backpropagation neural network (BPNN) and multilinear regression (MLR) models during calibration, and validation periods based on Nash-Sutcliffe efficiency (NSE), mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient (CC) indicators. According to the results of appraisal the hybrid GA-EANN model produced better outcomes (NSE = 0.9233/ 0.9018, MSE = 10.5195/ 9.7889 mg/L, RMSE = 3.2434/ 3.1287 mg/L, MAPE = 3.8032/ 3.0348 mg/L, and CC = 0.9609/ 0.9496) in calibration/ validation phases than BPNN and MLR models. In addition, the results indicate the better performance and suitability of the hybrid GA-EANN model with five input parameters in predicting the WQI for the study site. (c) 2021 Elsevier B.V. All rights reserved.
查看更多>>摘要：The motion of a ship, which has six degrees of freedom, is a complex nonlinear dynamic process with variable periodicity and chaotic characteristics. With the development of smart ships, modern high-precision equipment needs the help from high accuracy of ship motion (SHM) forecasting. Existing models will not easily be able to satisfy future accuracy requirements. Therefore, to improve the accuracy of SHM forecasts, by firstly determining the sequence features of SHM time series, a convolutional neural network (CNN) was used herein to extract automatically spatial feature vectors. Considering the variable-period characteristics of SHM time series, a gated recurrent unit (GRU) was used to learn the inherent time characteristics and to extract temporal feature vectors. The attention mechanism (AM) was developed to control the effect of feature vectors on the output to solve the problem of the contribution of feature vectors. Integrating the above methods, an SHM hybrid forecasting model, the SHM CNN-GRU-AM (SHM-C&G&A) model, was established. Secondly, in view of the difficulty of selecting the hyperparameters of a hybrid model, on account of the defects of the whale optimization algorithm (WOA), a normal cloud local search (NCLS) algorithm was developed. Exploiting the advantages of the normal cloud search (NCS) and the genetic algorithm (GA), a genetic random global search (GRGS) algorithm was developed. Then, a hybrid genetic cloud whale optimization algorithm (GCWOA) was developed and used to optimize the hyperparameters of the SHM-C&G&A model. Accordingly, a hybrid forecasting approach that integrates SHM-C&G&A and GCWOA was proposed; it is referred to as GCWOA-SHM-C&G&A. Finally, ship heave and pitch time series data are used to analyze an example to test the forecasting effectiveness of SHM-C&G&A and the optimization performance of GCWOA. The experimental results reveal that the proposed SHM-C&G&A model is more robust that the other models that are considered in this paper, and exhibits better nonlinear characteristics. The proposed GCWOA yields a better combination of hyperparameters than contrast algorithms in the forecasting process. (C) 2021 Elsevier B.V. All rights reserved.
NSTL
Elsevier