查看更多>>摘要：Evolutionary algorithms require large number of function evaluations to locate the global optimum, making it computationally prohibitive on dealing with expensive problems. Surrogate-based optimization methods have shown promising ability on accelerating the convergence speed. However, it is still a challenging work for surrogate-assisted methods to deal with high-dimensional expensive problems because it is hard to approximate the objective function in high-dimensional space. In this paper, a novel radial basis function surrogate model assisted evolutionary algorithm for high-dimensional expensive optimization problems (RSAEH) is proposed. Specifically, the proposed algorithm consists of local search part and surrogate-guided prescreening part. In the local search part, the local surrogate is built by radial basis function with the most promising training sample points, and the optima (or near-optima) is located by optimizer to conduct exact function evaluation. In the surrogate-guided prescreening part, the current best sample point is refined by using sequential quadratic programming, thus guide the mutation direction by using differential evolution operator, and promising offspring prescreened by surrogate model are evaluated using exact function evaluation. To validate the effectiveness of the proposed algorithm, it is tested on benchmark problems with dimension ranging from 30 to 100, as well as a real-world oil reservoir production optimization problem. The proposed algorithm achieved best optimization results on 16 benchmark functions among 21 benchmark function sets in comparison with other algorithms. The performance of RSAEH is competitive especially on 100-D benchmark functions. In addition, RSAEH also showed promising performance on a real-world oil reservoir production optimization problem with 160 variables, in comparison with several state-of-the-art algorithms.
查看更多>>摘要：Collaborative filtering (CF), the most widely used recommendation algorithm, has to face the sparsity and scalability problem. Some researchers proposed to select a representative set of real users called information core (IC) from all the real users, which is used as the candidate neighbor set in the CF to alleviate the scalability problem. However, the rating vectors of these real users that compose IC are usually sparse, which will negatively affect the recommendation accuracy. In this paper, a virtual information core (VIC) optimization algorithm is proposed based on clustering and evolutionary algorithms for CF recommendation (VICO-CEA). The problem of searching for VIC is modeled as a combinatorial optimization problem, and is solved offline by the proposed evolutionary algorithm. The VIC is a set of virtual core users, each of which is constructed by averaging out multiple real users. These virtual core users in the VIC are no longer sparse and are found by the evolutionary optimization, which will improve the recommendation accuracy and reduce the online recommendation time as the VIC is used as the candidate neighbor set in the CF. Meanwhile, to make offline optimization more efficient, two strategies are proposed. One is to design a simple similarity measure based on dimensionality reduction and clustering to save time in calculating similarities by reducing the dimensionality of users’ rating vectors. The other is to use dimensionality reduction and clustering to construct a smaller training set and validation set by reducing the dimensionality of items’ rating vectors. The experimental results show that VICO-CEA can not only significantly reduce the online recommendation time further but also improve the recommendation accuracy greatly compared to traditional CF and other information-core-based methods.
查看更多>>摘要：Edge computing, which provides lightweight cloud computing and storage capabilities at the edge of the network, has become a new computing paradigm. A key research challenge for edge computing is to design an efficient offloading strategy for offloading decision-making and resource allocation. Although many researches attempt to address this challenge, the traditional offloading strategies cannot adapt to complex environments, and the offloading strategies based on reinforcement learning require centralized control or the pursuit of the user's best interests, which is impractical. Individual users should rationally pursue benefits in order to create a high-quality offloading environment to obtain long-term benefits. In this paper, we first separate the offloading process into a two-step offloading framework, and reverse the order of solving offloading decision and resource allocation problems to reduce the dimensionality of the action and state space. We formulate the resource allocation as a Markov Decision Process (MDP) and use the Deep Deterministic Policy Gradient Algorithm (DDPG) to adjust load balancing of the edge server and reduce the transmission energy and delay, and then use the genetic algorithm (GA) to search for decisions and use Fully-Connected Network (FCN) to fit the decision-making process, thereby avoiding excessive response time caused by iteration. Simulation results show that compared with baseline methods, the proposed algorithm is more stable, flexible, adaptable and suitable for practical applications.
查看更多>>摘要：Many of decision-making and policy planning processes involve a time-series prediction problem and so this area has extensive literature including a great variety of time-series prediction tools and inferences systems. An important part of these is based on fuzzy sets. However, it is known that fuzzy sets may fail to satisfy or characterize the uncertainty of the data in a comprehensive manner because they cannot depict the neutrality degree of time-series. Another important and decisive deficiency of current inference systems is to based on the univariate structure. However, the time series dealt with in a prediction problem generally interact with other time series. Considering these issues, creating an inference system based on intuitionistic fuzzy sets and multivariate relationships for a time series prediction problem is a requirement even an obligation. With these regards, this study presents a multivariate intuitionistic fuzzy time-series definition and its prediction models and introduces a multivariate intuitionistic fuzzy inference system (M-IFIS). The basic novelty of the article can be expressed as the definition of a multivariate intuitionistic fuzzy time series, as well as the creation of a relevant analysis mechanism, first-time in the literature. Sigma-pi neural network is used as an inference tool in M-IFIS and membership and non-membership values and lagged crisp observations of multivariable time-series are used as inputs of it. In order to reveal the performance of the proposed system, Istanbul Stock Exchange (IEX) and Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) are analysed and the results are evaluated as comprehensive and comparative. All findings reveal the superiority M-IFIS in predictive accuracy.
查看更多>>摘要：Recently, Ghanbari and Mahdavi-Amiri (2011) have proposed a model for the bus terminal location problem. Here, we want to improve Ghanbari and Mahdavi-Amiri's model by defining three types of neighborhoods for each terminal. In our model, we consider a neighbor for each terminal, so the cost of service to stations in a neighborhood is individual. The proposed model is an NP-hard problem, so we suggest two algorithms based on the Tabu search and variable neighborhood search for solving it. Due to the Mann–Whitney and Dolan–Moré performance profiles, we use the recently proposed nonparametric statistical test to access the performance of numerical algorithms and demonstrate the efficiency of our proposed approach in comparison with other available methods.
查看更多>>摘要：The quality of datasets is one of the key factors that affect the accuracy of aerodynamic data models. For example, in the uniformly sampled Burgers’ dataset, insufficient high-speed data is overwhelmed by massive low-speed data. Predicting high-speed data is more difficult than predicting low-speed data, owing to the fact that the number of high-speed data is limited, i.e. the quality of the Burgers’ dataset is not satisfactory. To improve quality of datasets, traditional methods usually employ data resampling technology to produce enough data for the insufficient parts in the original datasets before modeling, which increases computational costs. Motivated by the mixtures of experts in natural language processing, we propose a multi-task learning (MTL) scheme in the field of aerodynamic data predictions to eliminate the need for data resampling. Our MTL is a datasets quality-adaptive learning scheme, which combines task allocation and aerodynamic characteristic learning together to disperse the pressure of an entire learning task. The task allocation divides a whole learning task into several independent subtasks, while the aerodynamic characteristic learning learns these subtasks simultaneously to achieve better precisions. Two experiments with poor quality datasets are conducted to verify the data quality-adaptivity of the MTL to datasets. The results show that the MTL whose subtasks are divided by the K-means is more accurate than fully connected networks (FCNs), generative adversarial networks (GANs) and radical basis function neural networks (RBFNNs) in poor quality datasets.
查看更多>>摘要：A problem related to the development of algorithms designed to find the structure of artificial neural network used for behavioural (black-box) modelling of selected dynamic processes has been addressed in this paper. The research has included four original proposals of algorithms dedicated to neural network architecture search. Algorithms have been based on well-known optimisation techniques such as evolutionary algorithms and gradient descent methods. In the presented research an artificial neural network of recurrent type has been used, whose architecture has been selected in an optimised way based on the above-mentioned algorithms. The optimality has been understood as achieving a trade-off between the size of the neural network and its accuracy in capturing the response of the mathematical model under which it has been learnt. During the optimisation, original specialised evolutionary operators have been proposed. The research involved an extended validation study based on data generated from a mathematical model of the fast processes occurring in a pressurised water nuclear reactor.
查看更多>>摘要：DevOps is an emerging paradigm that refer to a collaborative culture of development and operation teams aiming to develop the high quality software product. Software organizations are adopting DevOps culture for software development and easy maintenance instead of using traditional SDLC mechanism. To enter the production stage, in DevOps process, the software product have to pass through quality gates were the software are tested during development phase to meet the established targeted criteria. This indicates that the mechanism of testing in DevOps process is not straightforward, and to establish strong DevOps testing platform there is a need to explore more automated testing practices. Thus, using multivocal literature review approach, we have selected 39 studies and identify the 20 testing capabilities. Finally, the interpretive structure modeling (ISM) and fuzzy technique for order preference by similarity to ideal solution (fuzzy TOPSIS) were applied. The results shows that (C2, CCi=0.808; C6, CCi=0.720; and C3, CCi=0.705) are top ranked testing capabilities. Using analysis results, we develop a holistic structure of testing capabilities to show their inter-relationship with each other and their priorities to select the best testing capabilities for DevOps process.
查看更多>>摘要：The scarcity of resources requires a decrease in nonrenewable energy consumption, which progressively promotes the development of renewable energy due to its immense potential and environmental friendliness. Hence, the use of renewable energy technology is critical for realizing the economic effect, the environment effect and the social benefit unified. Generally, renewable energy technology selection is treated as a multiple criteria group decision making problem. However, decision makers are not allowed to express multiple preferences via personalized linguistic distribution assessments deliberating on diverse criteria in the existing approaches. This work proposes a multi-granular linguistic distribution-based group decision-making method by linking multi-granular linguistic distribution assessments and LINMAP (Linear Programming Technique for Multidimensional Analysis of Preference) method with a mathematical model that can simultaneously yield the credible weights of the considered criteria and prioritize the sequence of optimal renewable energy technologies. To this end, the linguistic distribution-based Hellinger distance measure and linguistic hierarchy-based multi-granular linguistic distribution transformation method are proposed. The decision framework is applied to a case study of power generation-based technology selection, generating reliable criteria weights and yielding acceptable outcomes based on collected assessments. Eventually, the sensitivity analysis and comparative analysis are conducted to verify the feasibility and practicability of our proposal. This flexible decision support technique is geared towards managers and strives to provide reference and inspiration for renewable energy technology selection.
查看更多>>摘要：Modern industrial fields utilize complex mechanical equipment and machinery, which are closely linked, and equipment faults are difficult to express. Therefore, fault diagnosis is important to ensure the safety of complex mechanical equipment in modern industries. Deep learning has achieved excellent results with recent fault diagnosis methods. At present, three common deep learning models (MLP, CNN, and RNN models) can achieve diagnosis rates close to 100% with original fault diagnosis data and a signal-to-noise ratio above 10 dB. However, we found that the diagnostic rate of these three models was completely incorrect when an adversarial sample with a signal-to-noise ratio noise greater than 10 dB was added to the original sample. We propose a GAN-based adversarial signal generative adversarial network (AdvSGAN) in this paper. We conduct experiments on the CWRU dataset and conclude that we can easily obtain adversarial noise and generate training samples through AdvSGAN. With the addition of adversarial data training, the diagnostic rate of the model on these adversarial samples increased from less than 5% to 98.69%, 97.38% and 96.94%. Hence, this method increases the reliability of our deep learning model.
NSTL
Elsevier