查看更多>>摘要：This paper aims to improve the explainability of autoencoder (AE) predictions by proposing two novel explanation methods based on the mean and epistemic uncertainty of log-likelihood estimates, which naturally arise from the probabilistic formulation of the AE, the Bayesian autoencoder (BAE). These formulations contrast the conventional post-hoc explanation methods for AEs, which incur additional modelling effort and implementations. We further extend the methods for sensor-based explanations, aggregating the explanations at the sensor level instead of the lower feature level. To evaluate the performance of explanation methods quantitatively, we test them on condition monitoring applications. Due to the lack of a common assessment of explanation methods, especially under covariate shift, we propose three evaluation metrics: (1) the G-mean of Spearman drift coefficients, (2) the G-mean of sensitivity-specificity of explanation ranking and (3) a sensor explanation quality index (SEQI) which combines the first two metrics, capturing the explanations’ abilities to measure the degree of monotonicity and to rank the sensors. Surprisingly, we observe that the explanations of BAE's predictions suffer from high correlation resulting in misleading explanations. This new observation cautions against trusting these explanations without further understanding of when they may fail. To alleviate this, a “Coalitional BAE” is proposed, inspired by agent-based system theory. The Coalitional BAE models each sensor independently and eliminates the correlation in explanations. Our comprehensive experiments on publicly available condition monitoring datasets demonstrate significant improvements of the Coalitional BAEs over the baseline Centralised AEs on the proposed metrics, visualised through critical difference diagrams.
查看更多>>摘要：Classifiers sometimes return a set of values of the class variable since there is not enough information to point to a single class value. These classifiers are known as imprecise classifiers. Decision Trees for Imprecise Classification were proposed and adapted to consider the error costs when classifying new instances. In this work, we present a new cost-sensitive Decision Tree for Imprecise Classification that considers the error costs by weighting instances, also considering such costs in the tree building process. Our proposed method uses the Nonparametric Predictive Inference Model, a nonparametric model that does not assume previous knowledge about the data, unlike previous imprecise probabilities models. We show that our proposal might give more informative predictions than the existing cost-sensitive Decision Tree for Imprecise Classification. Experimental results reveal that, in Imprecise Classification, our proposed cost-sensitive Decision Tree significantly outperforms the one proposed so far; even though the cost of erroneous classifications is higher with our proposal, it tends to provide more informative predictions.
查看更多>>摘要：Structural health monitoring (SHM) is a substantial research direction in structural engineering being scrutinized in recent years due to its significance in ensuring structural safety and reliability. The SHM using evolutionary computation has gained wide significance thanks to its computational capability and robust performance. By bearing in mind the necessity of efficient search mechanisms, a new algorithm merging the recent social engineering optimizer (SEO) and the particle swarm algorithm (PSO) is proposed in this work. The new algorithm, called the social engineering particle swarm optimization algorithm (SEPSO), combines the PSO population-based elitist-solution mechanism and the SEO two-solution attacker–defender paradigm to establish an effective global–local algorithm for solving complex engineering optimization problems. The SEPSO is developed, benchmarked, and compared with some state-of-the art algorithms using available test functions in the literature. Furthermore, the SEPSO is applied on selected mechanical design problems from the CEC2020 real-world constraint optimization competition in addition to comparison with the best-performed algorithms for benchmarking purpose. The SEPSO algorithm exhibits outstanding results when solving the benchmark functions and the real-world constraint optimization problems. Moreover, aiming to solve the complex problem of the SHM inverse problem, the SEPSO is employed. The SHM framework is presented and applied on the American Society of Civil Engineering (ASCE) frame structure based on an efficient fusion of objective function formulation. Several damage cases are tested using partial and noise-contaminated data. The proposed approach shows notable detectability and severity evaluation regardless the modal data malfunctions. Meanwhile, the proposed SEPSO can serve as an effective global–local search algorithm for solving real-world engineering problems.
Gomes Pereira de Lacerda M.Ludermir T.B.de Andrade Amorim Neto H.Buarque de Lima Neto F....
17页
查看更多>>摘要：We present an innovative step towards a parameterless out-of-the-box population size control for evolutionary and swarm-based algorithms for single objective bound constrained real-parameter numerical optimization. To the best of our knowledge, our approach is the first parameterless out-of-the-box parameter control for such a kind of technique. It is easy to implement and to use, since it does not require the adjustment of any parameter. The general idea is to increment the velocity of the population change if the best fitness stagnates, and decrement it otherwise. Then, in order to effectively change the population size, a mechanism of removal/addition of individuals inspired by the selection methods of evolutionary algorithms is executed. Our experimental results provide evidence that our controller is not only compatible with any evolutionary or swarm-based algorithm for single objective bound constrained real-parameter numerical optimization, but that it also performs well in many scenarios.
查看更多>>摘要：The deep learning algorithm has made great breakthroughs in optical image processing. Some deep learning algorithms require a large number of labeled samples for training. For PolSAR data sets, due to the influence of speckle noise and other factors, high-quality labeled data are limited. Therefore, it is meaningful to use deep learning algorithm to solve PolSAR classification problem in limited labeled dataset. This paper proposes a spatial feature-based convolutional neural network (SF-CNN). The network adopts a dual-branch CNN structure. Both of the two branches have the same structure and share parameters. SF-CNN can receive more than one sample as input. SF-CNN's special structure can expand the original training set by combining different samples, and alleviate the problem of insufficient labeled training data in PolSAR image classification tasks. When training, SF-CNN maps high-dimensional PolSAR image to low-dimensional feature space. In low-dimensional feature space, SF-CNN enhances the ability of network to extract discriminative features by maximizing or minimizing the distance between feature centers of different classes. In order to dig up the relationship between the samples, the test sample features are compared with every training sample feature when testing. Finally, labels of test samples are determined by the comparison result. The result of SF-CNN in PolSAR image classification task is better than that of standard CNN.
查看更多>>摘要：Developing a prognostic model to predict an asset's health condition is a maintenance strategy that increases asset availability and reliability through better maintenance scheduling. Therefore, developing reliable vehicle health predictive models is vital in the aerospace industry, especially considering a safety–critical system such as aircraft. However, one of the significant challenges faced in building reliable data-driven prognostic models is the imbalance dataset. Training machine-learning models using an imbalanced dataset causes classifiers to be biased towards the class with majority samples, resulting in poor predictive accuracy in data-driven models. This problem can become more challenging if the imbalance ratio is extreme and classes overlap. In this paper, a novel approach called Balanced Calibrated Hybrid Ensemble Technique (BACHE) is developed to tackle the severe imbalanced classification problem. The proposed method involves the combination of hybrid data sampling and ensemble-based learning. It uses a cascading balanced approach to transfer a class imbalance problem into a sub-problem by decomposing the original problem into a set of subproblems, each characterized by a reduced imbalance ratio. Then uses a calibrated boosting with a cost-sensitive decision tree to enhance recognition of hard-to-learn patterns, which improves the prediction of the extreme minority class. BACHE is evaluated using a real-world aircraft dataset with rare component replacement instances. Also, a comparative experiment of the proposed approach with other similar existing methods is conducted. The performance metrics used are precision, recall, G-mean, and an area under the curve. The final results show that the proposed model outperforms other similar methods. Also, it can attain an excellent performance on large, extremely imbalanced datasets.
查看更多>>摘要：In the apparel industry, fabric contributes to the high cost of raw materials, and thus an improvement in terms of a shorter used marker layout would improve the cost efficiency of this industry. The marker planning problem, also known as a 2D irregular cutting and packing problem in the apparel industry, focuses on optimizing the fabric resource by arranging a set of irregularly shaped clothing patterns on a sheet of fabric while preventing any overlap between the patterns, with the aim of finding the shortest length arrangement. Due to the irregular shapes of clothes, the solution time increases exponentially when more pieces are involved, making this problem become NP-hard or NP-complete. In this study, particle swarm optimization (PSO)-based heuristics were evaluated to address the above problem. The moving heuristic proposed by Tsao et al. (2020) acts as a placement strategy considering the order of the pattern and the degree of rotation. A pixel-based representation was used to handle the geometry of the pattern. PSO-based heuristics were developed by enhancing PSO performance with a local search, a genetic algorithm, and simulated annealing. Mixed-size order and a special case of the separated-size arrangement were also considered. The proposed algorithms were tested in an apparel company and compared with the well-known bottom-left fill heuristic approach to obtain competitive results with shorter fabric length and CPU time.
查看更多>>摘要：Large mass gatherings such as pilgrimages, protests, etc., often pose serious challenges for the crowd management personnel to maintain public safety and security especially in dense crowds. These challenges can be mitigated through estimating the number of attendees as well as localizing them in a particular crowded event, where existing research studies are yet to provide accurate information in an efficient manner. Therefore, in this paper, we propose a novel deep learning architecture namely LC-Net to precisely and efficiently locate as well as count the attendees in dense crowds using a crowd localization map. Here, we exploit the notions of residual layers and dilated convolution to improve both the accuracy and efficiency of our architecture. Besides, we propose a new data augmentation technique to resize the high-resolution training images based on crowd density that substantially boosts our localization accuracy. Rigorous experimental evaluation of our proposed LC-Net over four different public crowd datasets such as NWPU-Crowd, UCF-QNRF, ShanghaiTech-A, and ShanghaiTech-B shows a substantial performance improvement while using LC-Net in terms of precision and recall in most of the cases. The improvement eventually results in an improved F1 score in all cases compared to the state-of-the-art approaches. Further, we present a real implementation of our proposed approach using a client–server application. In the server, we execute the LC-Net model over the images captured in real-time using an IP Camera and then visualize the results in a graphical manner. This implementation demonstrates the applicability of our proposed approach in real cases.
查看更多>>摘要：Prediction methods have become a hot topic in intelligent decision making. Most of the existing prediction methods focus on the prediction accuracy and stability. As a second choice, accurate interval prediction can provide a relatively reliable reference in the sense of probability and provide help for assisting decision management. Therefore, we propose a novel interval prediction approach. Firstly, the decomposition method based on ensemble empirical mode decomposition (EEMD) is utilized to alleviate the complexity of the original time series, thereby generating a series of relatively smooth subseries. Secondly, a three-way clustering (TWC) algorithm is established by integrating sample entropy into probabilistic rough set, enriching the three-way clustering theory from the perspective of entropy. Thirdly, aiming at determining the optimal input dimensions of different neural networks, the feature selection technique based on phase space reconstruction (PSR) is constructed. Furthermore, an interval prediction system based on TWC is proposed to provide a new data-driven prediction method. Finally, the proposed approach is applied to predict the interval price of crude oil. On the one hand, the practicability of the constructed prediction approach is verified; on the other hand, it provides a new theoretical method for interval prediction of crude oil price. The experiment results show the proposed prediction approach can assist the decision-makers to make scientific and reasonable decisions.
查看更多>>摘要：The availability of a large amount of high-quality data is critical to the performance of machine-learning models. It is challenging to obtain a training dataset because data collection is costly and time-consuming. However, data scarcity can be overcome and an accurate model can be obtained if data from similar models are reused. In this paper, we propose an instance-based transfer learning method to obtain a more accurate model for situations with data scarcity. The proposed method uses a modified domain-adaptation technique to generate auxiliary target-domain data from source-domain data. Subsequently, useful data are selected from the auxiliary target-domain data to preclude the negative transfer that may leverage source-domain data to reduce the learning performance in the target domain. A modified domain-adversarial neural network was used to generate auxiliary target-domain data in the context of instance-based transfer learning. Particularly, the feature extractor and domain discriminator were trained to extract the domain-invariant features from the source and target domains, whereas the target generator was trained to generate auxiliary target-domain data using the domain-invariant features. Additionally, an influence function that can measure the influence of individual training samples on the learning performance was applied to identify useful data. Three case studies were conducted to validate the proposed method: a mathematical function example, drone blade metamodeling, and bearing fault diagnosis. The results of these case studies indicate a significant improvement in neural network prediction despite data scarcity.
NSTL
Elsevier