Search

The main aim of feature subset selection is to find the minimum number of required features to perform classification without affecting the accuracy. It is one of the useful real-world applications for different types of classification datasets. Different feature subsets may achieve similar classification accuracy, which can help the user to select the optimal features. There are two main objectives involved in selecting a feature subset: minimizing the number of features and maximizing the accuracy. However, most of the existing studies do not consider multiple feature subsets of the same size. In this paper, we have proposed an algorithm for multimodal multiobjective optimization based on differential evolution with respect to the feature subset selection problem. We have proposed the probability initialization method to identify the selected features with equal distribution in the search space. We have also proposed a niching technique to explore the search space and exploit the nearby solutions. Further, we have proposed a convergence archive to locate and store the optimal feature subsets. Exhaustive experimentation has been conducted on different datasets with varying characteristics to identify multiple feature subsets. We have also proposed an evaluation metric for the quantitative comparison of the proposed algorithm with the existing algorithms. Results have also been compared with existing algorithms in the objective space and in terms of classification accuracy, which shows the effectiveness of the proposed algorithm. • Multimodal multiobjective feature subset selection using differential evolution is explored in this paper. • Individuals initialized using Probability Based Initialization technique. • Balanced Niching Technique used to create and balance niches. • Stagnated Archive to identify the stagnated individuals and explore search space. • Multi Set Value metric for identifying the multiple feature subsets obtained. [ABSTRACT FROM AUTHOR]

Multiobjective evolutionary algorithms (MOEAs) have gained much attention due to their high effectiveness and efficiency in solving multiobjective optimization problems (MOPs). However, when solving MOPs, it is important but difficult to maintain a good balance of exploration and exploitation. In addition, some reference point based MOEAs with fixed reference points perform poorly on MOPs with irregular frontiers. Therefore, this paper proposes a new multistrategy multiobjective differential evolutionary (DE) algorithm, named RLMMDE. In RLMMDE, a multistrategy and multicrossover DE optimizer is utilized to alleviate the exploration and exploitation dilemma. An adaptive reference point activation mechanism based on RL is proposed to activate the adaptive adjustment of reference points. Moreover, a reference point adaptation method is proposed to improve the performance of RLMMDE on irregular frontier problems. Experimental results of RLMMDE tested on some benchmark test suites (i.e., ZDT, DTLZ, UF, WFG, and LSMOP) and two practical mixed-variable optimization problems show that the algorithm outperforms some advanced MOEAs. [ABSTRACT FROM AUTHOR]

Unmanned Aerial Vehicles (UAVs) are frequently utilized for transferring and delivering. Dual- or multi-quadrotor cooperative transportation schemes are required for heavy loads. In this work, focusing on the path planning for a dual-quadrotor cooperative transport system, a unique artificial potential field-A*(APF-A*) algorithm is studied to fulfill the safe and effective cooperative suspension transport. APF-A* algorithm proposed the strategy of host-wingman cooperative transportation based on the APF theory, and optimizes the heuristic function by introducing the concept of repulsive force to maintain the quadrotor flying at a safe distance from the obstacle. In addition, APF-A* algorithm promises to increase search efficiency while identifying the optimal path, and it can reduce the swing angle motion of loads to enhance the security of transportation. The coordinated flight control and path planning of quadrotor UAVs are simulated in AirSim environment and tested in flight experiments. The results demonstrate that the algorithm is capable to plan safe and reliable paths, meanwhile changing formation and avoid collision when the wingman is close to obstacles. In comparison to the regular A* algorithm, the APF-A* algorithm proposed in the paper performs better in path planning for dual-quadrotor cooperative transport system, reducing the search time and the number of search grids by nearly 60% and 53.75%, respectively. [ABSTRACT FROM AUTHOR]

This paper proposes two stochastic variance reduced gradient algorithms based on adaptive complex-valued limited-memory BFGS (ACL-BFGS) algorithm, which are established on the calculation of Wirtinger gradient for fully complex-valued dendritic neuron model (FCDNM). Basically, the explicit expression of Wirtinger gradient is derived to enable the quasi-Newton algorithm for the training of FCDNM. A stochastic variance reduction strategy is adopted such that the number of samples per batch is increased during the optimization process to ensure the effectiveness of the stochastic algorithm. The stochastic ACL-BFGS algorithm with variance reduction (SACL-BFGS) can be efficiently applied for the training of FCDNM. Furthermore, by generalizing the A3 accumulative multi-step quasi-Newton method, the A3-SACL-BFGS algorithm is presented to achieve fast convergence. The experimental results on some benchmark prediction and classification problems conclusively imply that the proposed SACL-BFGS and A3-SACL-BFGS algorithms for FCDNM are clearly superior to complex-valued backpropagation (CBP), CL-BFGS and ACL-BFGS algorithms. [ABSTRACT FROM AUTHOR]

An efficient convolution kernel could transform images into more expressive representations, which usually determines the quality of image classification models. Algorithms in the field of hyper-parameter optimization and neural architecture search could be applied to mining optimal kernels, while hyper-parameter optimization algorithms always require multiple training sessions to evaluate hyper-parameter configurations, resulting in a high time consumption. Recent gradient-based architecture search algorithms could deal with such problems within a single training session. However, there are currently no gradient-based NAS algorithms that could be used to efficiently optimize convolution kernels. This paper proposes a novel gradient-based algorithm, SoftStep, which could precisely and efficiently fine-tune the hyper-parameters of kernels. The proposed SoftStep is applied for the optimization of the convolution kernel's key hyper-parameters and generalized to more hyper-parameters in deep learning. In experiments, architecture search tasks on multiple challenging datasets are compared to various state-of-art hyper-parameter optimization and architecture search algorithms. The reported results show that SoftStep could mine optimal kernels far more efficiently than other SOTA methods. [ABSTRACT FROM AUTHOR]

With a small scale of prior information, including pairwise constraints and partial labels, semi-supervised clustering algorithms can improve clustering performance by imposing certain restrictions on the clustering results, and have accordingly attracted a lot of attention in recent years. Moreover, a large number of graph-based semi-supervised clustering algorithms have been proposed, since graphs are naturally suited to pairwise constraints. However, most graph-based approaches involve a two-step process and are not scalable to large-scale datasets due to their high time complexity. Additionally, they also encounter a serious problem of constraint violation. To deal with this issue, we propose a new non-parametric model to directly obtain the indicator matrix by minimizing the cut of the graph with a regularization term that corresponds to the balance of the clusters. To simultaneously support the two forms of prior information, our approach restricts the indicator matrix during optimization, while also ensuring that the clusters obtained by our approach do not violate any constraints. The computational complexity of the proposed approach is nearly linear to the data size n. Experiments show that the algorithm proposed in this paper achieves significant improvements in both clustering performance and efficiency. [ABSTRACT FROM AUTHOR]

Multi-scenario multi-objective optimization problems (MSMOPs) are a topic of considerable interest in the field of optimization. The MSMOP contains a set of multi-objective optimization problems arising from varying operating conditions, with the goal of determining a set of communal compromise solutions. However, there are few universal methods available for MSMOPs. This paper presents a general method incorporating transfer learning for MSMOPs. First, a multi-scenario ensemble framework that transfers knowledge between scenarios is developed to combine arbitrary multi-objective evolutionary algorithms, where a scenario-based comprehensive evaluation indicator is developed for combination with base learners. Then, an adaptive decomposition-based multi-objective evolutionary algorithm with a bi-layer selection (EADaBS) is proposed and embedded within the framework as a base learner. EADaBS incorporates an adaptive fitness assignment in its first layer to facilitate exploration, and density measurement in its second layer to ensure exploitation. A rebalancing operator is also designed to aid the population towards the Pareto front. Finally, a multitude of experiments is conducted to verify the effectiveness and efficiency of the proposed algorithms. Two three-scenario multi-objective optimization problems are designed and utilized as test problems. The experimental results clearly demonstrate that the proposed framework outperforms existing state-of-the-art algorithms. In conclusion, this research provides new insights into the solution of MSMOPs. [ABSTRACT FROM AUTHOR]

Multi-level image thresholding is a common approach to image segmentation where an image is divided into several regions based on its histogram. Otsu's method is the most popular method for this purpose, and is based on seeking for threshold values that maximise the between-class variance. This requires an exhaustive search to find the optimal set of threshold values, making image thresholding a time-consuming process. This is especially the case with increasing numbers of thresholds since, due to the curse of dimensionality, the search space enlarges exponentially with the number of thresholds. Population-based metaheuristic algorithms are efficient and effective problem-independent methods to tackle hard optimisation problems. Over the years, a variety of such algorithms, often based on bio-inspired paradigms, have been proposed. In this paper, we formulate multi-level image thresholding as an optimisation problem and perform an extensive evaluation of 23 population-based metaheuristics, including both state-of-the-art and recently introduced algorithms, for this purpose. We benchmark the algorithms on a set of commonly used images and based on various measures, including objective function value, peak signal-to-noise ratio, feature similarity index, and structural similarity index. In addition, we carry out a stability analysis as well as a statistical analysis to judge if there are significant differences between algorithms. Our experimental results indicate that recently introduced algorithms do not necessarily achieve acceptable performance in multi-level image thresholding, while some established algorithms are demonstrated to work better. [ABSTRACT FROM AUTHOR]

This paper addresses the problem classification of instances/questions based on the opinions (classes) provided by anonymous agents. The solution aggregates the agents' classifications, aiming to obtain as close as possible to an unknown correct classification. However, the agents' fields or domains of competence and their levels of expertise are unknown and can vary extensively. Many popular classification algorithms address such a problem by following a "wisdom-of-the-crowd" approach while using different voting methods and expectation–maximization techniques. These algorithms lead to correct classifications when the majority of the agents are experts, thus classifying the instances correctly. However, they often result in erroneous classification when only a small subset of the agents are indeed correct. Moreover, these algorithms often assume a fixed set of classes for all instances. This study presents a fast (one-pass) classification algorithm that can estimate the unknown agents' expertise level and aggregates their classifications accordingly, even when these are obtained from different questionnaires; thus, when the instances are not necessarily classified to a fixed set of classes. The proposed algorithm finds the experts and the nonexpert agents for each question by analyzing the distance between them. The algorithm identifies the expert agents for each instance and then classifies them accordingly. The suggested algorithm is validated and compared against known methods by using both simulated datasets and real-world datasets collected from various sources. The obtained results clearly demonstrate the effectiveness and advantages of the proposed method. [ABSTRACT FROM AUTHOR]

In this paper, we study the problem of detecting maximum k -durable structures on temporal graphs, which can be used to mine and analyze more knowledge behind the temporal graphs. We first prove that this problem is NP-complete and hard to approximate. Next, we propose an efficient algorithm to detect maximum k -durable structures. The algorithm accelerates the detection process by using an auxiliary graph and several well-designed pruning strategies. Massive experiments on five large temporal social networks demonstrate that our algorithm can save 2–4 orders of magnitude number of recursive invocation and is at least 30 × faster than the baseline algorithm. [ABSTRACT FROM AUTHOR]

In this paper, we propose an intelligent scheme and design a spatial information sampling algorithm (SIS) to achieve a balance between exploitation and exploration more efficiently. In SIS, by creatively using a chaotic map, we divide the population of the algorithm into external and internal parts and activate the corresponding exploitation or exploration operations according to the location where the optimal individuals appear in the population. The intelligent scheme makes full use of the spatial information in the solution space where the algorithm's population is located. It also gives the algorithm the authority to start exploitation or exploration operations autonomously based on the information in the space, which increases the flexibility of the algorithm in complex optimization problems. This novel scheme not only further promotes the intelligence of the algorithm in a simple architecture but also creates a new way of balancing exploitation and exploration in the optimization process. We test the SIS on IEEE CEC2017, IEEE CEC2011, an artificial neural model training problem, and a position optimization problem of wave energy converters. When compared to other state-of-the-art meta-heuristics, the results of the Wilcoxon signed-rank test, Wilcoxon rank-sum test, and Friedman test revealed that SIS possesses superior solutions in terms of global optimality, avoidance of local minima, and solution quality. The source code of SIS is released at https://toyamaailab.github.io/. • This paper proposes a new scheme to balance the exploitation and exploration. • The new scheme enables SIS to independently perceive, analyze, and take action. • SIS uses chaotic map to construct a population to realize an intelligent scheme. • SIS has high flexibility in dealing with different optimization problems. • The simulation results show that SIS is superior to the state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

The Successor Feature framework for Reinforcement Learning algorithms improves task transfer by decomposing the learned state–action value function. The decomposition involves two components, one that captures future-expected state features and the other that models the task-related reward structure. However, successful transfer between tasks depends heavily on how the reward function changes, possibly leading to failure of the original Successor Feature formulation. This paper proposes the Dynamic Successor Feature framework, DynSF, by extending the mathematical formulation of the original Successor Feature framework to center around a learned state-transition model. Under this formulation, the state-transition model dynamically induces the acting policy. The flexibility of DynSF also extends to the architecture, requiring only a state-transition model and a small vector of parameters. This architecture provides immense flexibility in the choice of the model used to learn the state-transition model. The DynSF framework is evaluated and compared to other baseline algorithms through several experiments in a continuous grid world environment, a robotic Reacher, and pixels in the Doom environment. • DynSF enables a state-transition model to be used in Reinforcement Learning. • Policy and discount factors can be parameterized dynamically. • Different supervised ML algorithms can be used to learn the state transition model. • DynSF can be used for guided exploration and transfer learning. • DynSF performs better during task transfer than the Successor Feature framework. [ABSTRACT FROM AUTHOR]

In heterogeneous computing systems, excellent task scheduling algorithms can shorten the task completion time and improve system parallelism. With the large-scale deployment of edge computing, the task scheduling algorithm in heterogeneous edge computing servers has become a critical factor in improving the overall system performance. This paper proposes a new task scheduling algorithm called the edge cover scheduling algorithm (ECSA), which schedules tasks based on the edge cover queue of the directed acyclic graph (DAG) for heterogeneous computing systems. Based on the estimation of distribution algorithm (EDA) and the graph random walk algorithm, the ECSA generates an edge cover queue from DAG. Then, the ECSA uses the heuristics greedy method with low time and computational complexity to allocate the edge cover queue to processors. Theoretical analysis and simulation results on random DAGs and real-world DAGs show that the ECSA can achieve better scheduling results in terms of makespan, the schedule length ratio (SLR), efficiency, and frequency of best results with low time and computational complexity. • A scheduling algorithm for HCSs based on the edge cover queue. • The edge cover queue explores a hybrid scheduling strategy of vertex and edge greed. • Generating edge cover queue based on the EDA and the graph random walk algorithm. • The strategy of a single population to simplify EDA. [ABSTRACT FROM AUTHOR]

In this paper, a dynamic multiobjective evolutionary algorithm (DMOEA) with an adaptive response mechanism selection strategy is proposed to address the shortcoming that a single response mechanism is suitable only for solving a certain type of dynamic multiobjective optimization problem. The proposed algorithm combines an adaptive response mechanism selection (ARMS) strategy and a multiobjective evolutionary algorithm based on decomposition (MOEA/D), and it is denoted as the MOEA/D-ARMS. Unlike the existing approaches, the ARMS strategy can adaptively select effective response mechanisms from the response mechanism pool based on the recent performance of each response mechanism. Four representative response mechanisms are selected to form the response mechanism pool. An overall evaluation strategy that assigns rewards to the response mechanism is adopted, and a probability-based method that is used to decide which response mechanism can be used to generate a new solution is employed. The proposed MOEA/D-ARMS algorithm is tested on two groups of test instances and compared with the decomposition-based and dominance-based DMOEAs. The results of the proposed MOEA/D-ARMS algorithm are superior to the compared algorithms, demonstrating its effectiveness. • This paper proposed an adaptive response mechanism selection (ARMS) framework and a dynamic MOEA. • The ARMS framework can adaptively select the most effective response mechanisms based on their recent performance. • An overall evaluation strategy that assigned rewards to the response mechanism was adopted. • The simple ARMS framework can easily incorporate more response mechanisms. • The statistical results clearly demonstrated that MOEA/D-ARMS were superior to the compared algorithms. [ABSTRACT FROM AUTHOR]

The segmentation of tunnel boring machine (TBM) time series plays a crucial role in analyzing TBM operating statuses and mining potential information from the collected signals. In this paper, a novel algorithm named SDP–RLR is proposed to segment multivariate TBM time series with outliers caused by harsh operating environments and changeable tunneling statuses. In this algorithm, the 3 σ rule is extended to externally studentized residuals of a linear regression and used to identify the outliers in each segment of the input time series. An outlier removal penalty term is added to the segment errors to avoid regarding borderline data points as outliers. As a segmentation optimization algorithm, dynamic programming (DP) is improved to scalable dynamic programming (SDP) in this paper by combining time series in two stages to reduce computational costs. In the first stage, consecutive time points are integrated to reduce the calculations required for DP, while the second stage refines the segmentation results obtained from the first stage. Experiments are conducted on synthetic datasets to evaluate the performance of the SDP–RLR algorithm, and a complete TBM time series in terms of three key variables validates its effectiveness for segmentation and outlier detection tasks. The SDP–RLR algorithm can segment a TBM time series into four different statuses, which assists in analyzing the operating statuses of TBMs. In addition, comparative experiments indicate that the proposed algorithm can handle the segmentation of complex TBM time series when they cannot be successfully processed using the segmentation approach after outlier detection. [ABSTRACT FROM AUTHOR]

Flexible job shop scheduling problem (FJSP) has attracted many research interests, in particular for meta-heuristic algorithm (MA) developers due to the superior optimization performance. Dragonfly algorithm (DA) is one of recent and popular MA approaches. However, it is inevitable for DA to be trapped into local optima, especially when dealing with the complex large-scale flexible job shop scheduling problem (LSFJSP). In this paper, an improved DA, adopting a dynamic opposite learning (DOL) strategy, is proposed (namely DOLDA) to solve the LSFJSP. DOL strategy is embedded into the population initialization stage and the generation jumping stage to raise the search ability of DA. This paper uses a T-test to testify whether there are differences between the proposed algorithm and other comparison algorithms, comparison results indicate that DOLDA shows noticeable differences with other algorithms that explain the effectiveness and innovation of the proposed algorithm. The jump rate is an important parameter that determines the probability of algorithms escaping from the algorithms' local optimal solution. This paper also considers the jump rate analysis experiments to maximize the power of the DOLDA. 28 test functions from CEC 2013 and CEC 2014 are applied to verify the performance of DOLDA, test results reveal that DOLDA owns strong search ability in coping with almost all test functions. The DOLDA also is applied to solve 15 LSFJSP instances generated by Brandimate rule, the results obtained show that DOLDA can efficiently achieve a better solution on the LSFJSP compare to compared algorithms. [ABSTRACT FROM AUTHOR]

Decision-making is one of the important deals in knowledge transfer and it can be useful for the multi-source domains. Meanwhile, the existing knowledge transfer schemes do not use privacy-preserving techniques for preserving security. This can be a problem for critical domains like financial market forecasting as the misuse of security can lead to legal and financial implications. In recent years, cloud services have revolutionized various technological applications. Cloud computing has become more popular with digital technologies as it provides uninterrupted services like transmission, storage, and intensive computing of data. The architecture of the cloud is also cost-efficient. Besides, various promising services from the cloud, some challenges need to be addressed to secure the privacy of the cloud users as millions of users access its services. Privacy preservation is an important aspect in the field of data mining, and the necessity of securing important data in the cloud from hackers is on the rise. Privacy-preserving data mining algorithms have been analyzed over recent years to provide sufficient solutions for securing the privacy of the data in the cloud. This paper plans to introduce a new hybrid meta-heuristic concept for developing a privacy preservation strategy towards business data under the cloud sector. The main objective of this paper is to design a new hybrid red deer-bird swarm algorithm (RD-BSA) to ensure higher convergence and since the use of control parameters over the solution generation is minimized. The proposed privacy preservation scheme on three financial databases is evaluated with the performance against the existing privacy preservation schemes. Different analyses like statistical, key sensitivity, Known-Plaintext Attack (KPA), and Chosen-Plaintext Attack (CPA) are used for evaluating the efficiency of the algorithm. The comparative analysis of the proposed model over the conventional models demonstrates its effective performance via diverse analysis. [ABSTRACT FROM AUTHOR]

The Chimp Optimization Algorithm (ChOA) is a hunting-based model and can be utilized as a set of optimization rules to tackle optimization problems. Due to agents' insufficient diversity in some complex problems, this algorithm is sometimes exposed to local optima stagnation. This paper introduces a Dynamic Lévy Flight (DLF) technique to smoothly and gradually transit the search agents from the exploration phase to the exploitation phase. To investigate the efficiency of the DLFChOA, this paper evaluates the performance of DLFChOA on twenty-three standard benchmark functions, twenty challenging functions of CEC-2005, ten suit tests of IEEE CEC06-2019, and twelve real-world optimization problems. The results are compared to benchmark optimization algorithms, including CMA-ES, SHADE, ChOA, HGSO, LGWO and ALEP (as the best benchmark Lévy-based algorithms), and eighteen state-of-the-art algorithms (as the winners of the CEC2019, the GECCO2019, and the SEMCCO2019). Among forty-three numerical test functions, DLFChOA and CMA-ES gain the first and second rank with thirty and eleven best results. In the 100-digit challenge, jDE100 with a score of 100 provides the best results, followed by DISHchain1e+12, and DLFChOA with a score of 85.68 is ranked fifth among eighteen state-of-the-art algorithms achieved the best score in seven out of ten problems. Finally, DLFChOA and CMA-ES respectively gain the best results in five and four real-world engineering problems. [ABSTRACT FROM AUTHOR]

The collaborative annealing power k-means++ (CAPKM++) clustering algorithm has been recently proposed based on multiple modules by minimizing annealed power-mean functions. This paper presents an upgraded version of CAPKM++ called CAPKM++2.0. Different from CAPKM++ where the anchor points of surrogate functions for majorizing the power-mean functions are re-initialized and minimized repeatedly after annealing, CAPKM++2.0 re-initializes the weights of the majorization function during annealing. In addition, unlike CAPKM++ that minimizes the majorization function of the power-mean sum, CAPKM++2.0 adds an inner loop to minimize the power-mean sum iteratively and locally at every annealing step. Ablation study results are discussed to justify the adoption of the power-mean and the collaboration of multiple modules. Experimental results on sixteen benchmark datasets are elaborated to demonstrate the superior clustering performance of the upgraded algorithm compared with its predecessor and six other mainstream algorithms in terms of cluster validity indices and algorithmic complexities. [ABSTRACT FROM AUTHOR]

Automated theorem prover (ATP) for first-order logic (FOL), as a significant inference engine, is one of the hot research areas in the field of knowledge representation and automated reasoning. E prover, as one of the leading ATPs, has made a significant contribution to the development of theorem provers for FOL, particularly equality handling, after more than two decades of development. However, there are still a large number of problems in the TPTP problem library, the benchmark problem library for ATPs, that E has yet to solve. The standard contradiction separation (S-CS) rule is an inference method introduced recently that can handle multiple clauses in a synergized way and has a few distinctive features which complements to the calculus of E. Binary clauses, on the other hand, are widely utilized in the automated deduction process for FOL because they have a minimal number of literals (typically only two literals), few symbols, and high manipulability. As a result, it is feasible to improve a prover's deduction capability by reusing binary clause. In this paper, a binary clause reusing algorithm based on the S-CS rule is firstly proposed, which is then incorporated into E with the objective to enhance E's performance, resulting in an extended E prover. According to experimental findings, the performance of the extended E prover not only outperforms E itself in a variety of aspects, but also solves 18 problems with rating of 1 in the TPTP library, meaning that none of the existing ATPs are able to resolve them. [ABSTRACT FROM AUTHOR]

Feature selection is an important preprocessing technique in the fields of data mining and machine learning. With the promotion of privacy protection awareness, recently it becomes a very practical and challenging issue to select high-quality feature subsets while ensuring the privacy of all participants. However, there is a lack of research results on this issue, i.e., feature selection under privacy protection. Aiming at the issue, this paper proposes a federated feature selection framework for the first time. In the framework, inspiring by the idea of federated learning, a credible third participant is introduced to process and integrate optimal feature subsets from multiple participants. On the basis of the framework, a federated evolutionary feature selection algorithm based on particle swarm optimization is proposed to effectively solve feature selection problems with multiple participants under privacy protection. Two new operators satisfying the requirement of privacy protection, i.e., the feature assembling strategy with multi-participant cooperation and the swarm initialization strategy guided by assembling solution, are designed to improve the ability of the proposed algorithm. Compared with several typical assembling feature selection algorithms on 15 data sets, experimental results show that the proposed algorithm can significantly improve the classification accuracy of the feature subset selected by each participant, while protecting the privacy of data. [ABSTRACT FROM AUTHOR]

Tensor ring (TR) decomposition, which factorizes a tensor into a sequence of cyclically interconnected third-order TR factors, is a powerful tool to capture the global low-rankness of high-dimensional data. However, the understanding of the physical interpretation of TR factors is not clear. In this paper, we first empirically discover the physical interpretation of TR factors in the gradient domain (termed as gradient factors) and then give the theoretical justification. Based on the interpretable gradient factors, we suggest a TR decomposition-based model with interpretable gradient factors regularization (TR-GFR) for tensor completion. To be specific, we consider the low-rankness and transformed sparsity priors of gradient factors to boost the performance and robustness of TR decomposition-based model. In addition, we develop an effective proximal alternating minimization algorithm to solve the proposed model. Numerical experiments validate that the proposed TR-GFR is superior to the compared state-of-the-art methods in terms of PSNR and SSIM values and more robust with TR rank. [ABSTRACT FROM AUTHOR]

Clustering has grown to be a research focus in recent years owing to the challenges of labeling massive collected data. Recent advances such as the emotional preference and migratory behavior clustering model jointly utilized the biogeography-based idea and inertia theorem for handling data clustering tasks. However, as with most clustering methods, the initialization for the solution is often purely random. Moreover, the search results vulnerably converge to local optima due to favorable exploration but short exploitation, which requires an additional process to balance them effectively. Inspired by the differential evolution (DE) strategy and the opposition-based learning (OBL) method, the paper proposed a novelty opposition-based differential evolution clustering algorithm for emotional preference and migratory behavior optimization named OBDE-EPMC. Specifically, we first utilize the OBL method to initialize the population to obtain a closer initial individual to the optimal solution. Thereafter, to ensure population diversity during the optimization, we applied the differential evolution (DE) strategy to expand the scope of the solution and the opposition-based learning method to generate an opposite one. Lastly, the historical solution scheme is added to retain the excellent solution within the population. In addition, the theoretical analysis and convergence property proof of OBDE-EPMC are presented. Numerous experiments were performed to compare the proposed OBDE-EPMC with the other seven clustering algorithms on a host of standard datasets. And several standards demonstrate the rationality and effectiveness of our proposed OBDE-EPMC algorithm. [ABSTRACT FROM AUTHOR]

Multi-Agent Path Finding (MAPF) aims to find a set of conflict-free paths for multiple agents on a given graph from parking locations to specified goal locations while optimizing related costs. Currently, existing MAPF studies often consider the simplified problem setup where each agent does not return to its parking location after completing its task on the underlying graph with uniform edge costs. Nevertheless, within some real-word scenarios such as the Unmanned Aircraft System (UAS), agents are situated in the underlying graph with non-uniform edge costs. These agents are required to travel from their respective parking locations to complete tasks, and then return without conflicts. Therefore, this paper explores a new version of MAPF, formally called Multi-Agent Path Finding with Heterogeneous edges and Roundtrips (MAPF-HR). In this version, all agents are engaged in completing tasks by navigating their respective conflict-free paths with roundtrips on the graph with heterogeneous edges. This paper investigates a novel algorithm for this problem, called Improved Conflict-Based Search (CBS) with Helpful Bypass (ICBS-HB), which improves the CBS framework by utilizing the scheme of bypassing conflicts during path finding execution. With extensive experiments on MAPF benchmark maps, it shows that ICBS-HB outperforms the state-of-the-art algorithms for dealing with MAPF-HR. [ABSTRACT FROM AUTHOR]

The recently introduced Marine Predators Algorithm (MPA) exhibits competitive performance in solving optimization problems. However, it often prematurely converges due to an imbalance between its exploration and exploitation capabilities. Therefore, in this paper, an improved MPA variant using a proposed Local Escaping Operator (LEO) is introduced. In the MPA, solution candidates are replaced by better candidates from the previous iteration, mimicking a memory of already visited prey-abundant areas. This indicates a weak inter-dependence between the candidates in the population and possible acceptance of new solution candidates created outside the algorithm without damaging the optimization process. Consequently, in the proposed approach, the worst candidates are replaced with solutions created by the LEO. The LEO is based on representative solutions, taking into account the relationship between predators and the characteristics of their population. The approach is experimentally compared with the state-of-the-art meta-heuristics on 82 test functions, including IEEE Congress on Evolutionary Computation test suites (CEC'14 and CEC'17) and three engineering problems. The results show the superiority of the LEO-MPA over the MPA and recent algorithms. Furthermore, in this paper, the suitability of the hybridization of meta-heuristics with the LEO is discussed and successful attempts with the best algorithms are shown. • A local escaping operator based on representative solutions is proposed. • Hybridization of the MPA algorithm with the LEO is introduced. • Introduced LEO-MPA is compared with 20 state-of-the-art meta-heuristics. • Suitability of the LEO to improve other meta-heuristics is investigated. [ABSTRACT FROM AUTHOR]

In the fields such as computer science, intelligent transport systems and logistics scheduling, many problems can be reflected as the variants of traveling salesman problem (TSP). In recent years, although there are many achievements and developments on the study of intelligent optimization algorithms for TSP and its variants, the challenge and problem are still existed: while the scale of problem is large, there is the dimension disaster problem; the optimization algorithms for TSP variants are easy to fall into the local optimum. Multiple balanced traveling salesmen problem (MBTSP) is a variant of TSP, it can be used in the fields such as optimizing multiple gas turbine engines. Since MBTSP is NP-hard problem, many intelligent optimization algorithms, such as genetic algorithm and ITÖ algorithm, have been used to solve it. However, while the scale of MBTSP is large, the traditional algorithms are easy to fall into local optimum. Aiming at the problem, this paper extends the scale of MBTSP to large scale scenes, and proposes a novel ITÖ algorithm (NITÖ) based on crossover operator and local search for large scale MBTSP. For NITÖ algorithm, the drift operator and volatility operator of ITÖ algorithm are redesigned, which are carried out by improved crossover operator and local search. The extensive experiments show that NITÖ can demonstrate better solution quality than the compared state-of-the-art algorithms. • Paper extends scale of multiple balanced traveling salesmen problem (MBTSP). • The large-scale optimization for MBTSP is firstly studied. • A novel ITÖ algorithm (NITÖ) is proposed for large scale MBTSP. • Local search is integrated into NITÖ algorithm for optimization. • Experiments show the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Closed high utility itemsets (CHUIs) and maximal high utility itemsets (MaxHUIs) are two important concise representations of HUIs. Discovering these itemsets is important because they are lossless and compact, i.e., they provide a concise summary of all HUIs that can be orders of magnitude smaller. In addition, it can be more efficient to extract these representations than it would be to extract all HUIs. Mining the concise representations of HUIs is also an important step toward the generation of nonredundant high utility association rules that can reveal meaningful information to decision-makers. However, although several algorithms have been designed to mine these representations, such as EFIM-Closed, HMiner-Closed, and CHUI-Miner(Max), they have long runtimes, high memory usage, and scalability issues, especially for dense and large datasets. To address this issue, this paper proposes two efficient algorithms named C-HUIM and MaxC-HUIM for mining CHUIs, and simultaneously mining both CHUIs and MaxHUIs, respectively. These algorithms use a novel weak upper bound on the utility, which is strictly tighter than traditional upper bounds, and a corresponding pruning strategy called SPWUB to quickly eliminate low utility itemsets. The algorithms also include two novel search space reduction strategies named PS N o n C H U B and LPS N o n C H U B. The PS N o n C H U B strategy only requires checking the inclusion relationship among a small number of itemsets, while LPS N o n C H U B does not perform any inclusion check. In addition, the algorithms adopt a structure named MPUN-list to efficiently store and calculate information about each itemset's utility and support. Experimental results show that the proposed algorithms can be more than 100 times faster, are more memory efficient, and have better scalability than the state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Domain adaptation aims to alleviate the shift between training and test distribution, where the DA theory is crucial in understanding the success of domain adaptation algorithms. In this paper, we reveal the incoherence between the empirical domain adversarial training and its generally assumed theoretical counterpart based on H -divergence. Concretely, we find that H -divergence is not equivalent to Jensen–Shannon divergence, the optimization objective in domain adversarial training. To this end, we establish a new theoretical framework by directly proving the upper and lower target risk bounds based on the joint distributional Jensen–Shannon divergence. We further derive bidirectional upper bounds for marginal and conditional shifts. Our framework exhibits inherent flexibility for different transfer learning problems, which is usable for various scenarios. From an algorithmic perspective, our theory enables a generic guideline of the unified principles of semantic conditional matching, feature marginal matching, and label marginal shift correction. We employ algorithms for each principle and empirically validate the benefits of our framework. [ABSTRACT FROM AUTHOR]

Label distribution learning, where each sample is associated with a distribution of description degree, suffers from the curse of dimensionality like other traditional learning paradigms. Feature selection as a pre-processing technique is commonly used to reduce the dimension of data. Currently, label distribution feature selection focuses on exploiting label correlations under the assumption that all samples share the same label correlations. However, the corresponding label correlations for different groups of samples may tend to be different in real-world tasks. To tackle the issue, this paper presents a novel approach named label distribution feature selection with feature weights fusion and local label correlations. First, instances are separated into several clusters regarded as local samples. Then, a two-strategy approach is presented based on the local samples. For the first, the feature weights obtained from all local samples are fused into uniform feature significance, which can effectively improve the feature discrimination. For the second, to reflect the influence of label correlations locally, a local correlation matrix is encoded via the label space of local samples as additional features. Subsequently, the feature weights for this strategy are generated from a new feature space composed of additional and original features. Finally, a subset of optimal and relevant features is selected from the top-ranked features based on feature weights obtained from the above two strategies. Extensive experiments on fifteen benchmark datasets verify that the proposed algorithm is superior to five state-of-the-art approaches in terms of six evaluation metrics. [ABSTRACT FROM AUTHOR]

The combination of Machine Learning (ML) algorithms is a solution for constructing stronger predictors than a single one. However, some approximations suggest that combining unstable algorithms provides better results than combining stable algorithms. For instance, Generative ensembles, based on re-sampling techniques, have demonstrated high performance by fusing the information of unstable base learners. Random Forest and Gradient Boosting are two well-known examples, both combining Decision Trees and providing better predictions than those obtained using a single tree. However, such successful results have not been achieved by assembling stable algorithms. This paper introduces the notion of limited learner and a new ensemble general framework called Minimally Overfitted Ensemble (MOE), a re-sampling-based ensemble approach that constructs slightly overfitted-based learners. The proposed framework works well with stable and unstable base algorithms, thanks to a Weighted RAndom Bootstrap (WRAB) sampling that provides the necessary diversity for the stable base algorithms. A hyperparameter analysis of the proposal is carried out on artificial data. Besides, its performance is evaluated on real datasets against well-known ML methods. The results confirm that the MOE framework works successfully using stable and unstable base algorithms, improving in most cases the predictive ability of single ML models and other ensemble methods. • New ensemble general framework. • A novel concept of ensemble learners is introduced. • Minimal overfitted models and limited learners are defined. • The framework works with any base machine learning training algorithm. • The proposal compares successfully to well known ensemble approaches. [ABSTRACT FROM AUTHOR]

Marine predators algorithm (MPA) is a newly developed swarm-based meta-heuristic algorithm that is inspired by the foraging strategies of marine predators: Lévy, Brownian motions and optimal encounter rate strategies, and has shown competitive performance with other state-of-the-art algorithms. Despite the effectiveness of original MPA and existing improved MPA in many real applications, the diversity of the population and the solution accuracy of MPA still deserve to be improved. To this end, an enhanced marine predators algorithm with the neighborhood-based learning strategy and the adaptive population size strategy is proposed in this paper, which is termed NMPA. Also, the superiority of the proposed NMPA is verified by being compared with the original MPA, other improved MPA and popular nature-inspired optimization algorithms on the well-known CEC 2017 and CEC 2020 test suites as well as three engineering design problems, respectively. Furthermore, the practicability of NMPA is illustrated by solving a real-world application of approximate developable surface modeling. By taking the maximization of developability degree of the developable surface as the evaluation criterion, the model for interpolating two given spatial curves to design an approximate developable surface is established. Then, the proposed NMPA is used to solve the optimization model, and the approximate developable surfaces with the maximum developability degree can be obtained. The experimental results show the superior performance of NMPA over other algorithms in terms of precision, convergence rate and solution quality. To sum up, NMPA is potentially a competitive and excellent meta-heuristic algorithm for constructing approximate developable surface and solving engineering optimization problems. • An enhanced marine predators algorithm termed NMPA is proposed. • The proposed NMPA is tested on CEC2017 and CEC2020 test suites as well as 3 engineering design problems. • The superiority and validity of NMPA is verified by comparing with multiple intelligent algorithms. • Optimization model for constructing developable surface through two given space curves is established. • The NMPA is applied to the solutions of the established optimization model. [ABSTRACT FROM AUTHOR]

AIOps(Artificial Intelligence Operations) is emerging as one of the most important technology in industrial automation, and accurate time series prediction plays a crucial role in it. However, due to the non-linear and non-stationary characteristics of Ops(operations) data, traditional time series forecasting models can not effectively extract good enough sequence data features and result in poor forecasting accuracy. To fully extract the Ops data information and construct a robust prediction model, this paper proposes a hybrid model based on recursive empirical mode decomposition (REMD) and memory wavelet neural network (MWNN) to improve the forecasting accuracy of Ops data. In REMD-MWNN, we first use REMD to decompose the Ops data into multiple intrinsic modal functions (IMF) at different time scales. Then, to make full use of the historical information of the Ops data and reduce the running time, we designed a new memory recurrent neural network, namely MWNN. Next, use MWNN to predict multiple inherent IMFs respectively to obtain the predicted value of the corresponding subsequence. Finally, the final prediction result is obtained by reconstructing the prediction value of each subsequence. In the comprehensive experiment, we selected the Ops data of an open education platform. The experimental results show that, compared with other algorithms, the model proposed in this paper is highly competitive in predicting future changes and capturing the evolution mode of hidden factors. The experimental data involved in this paper can be downloaded from this website: https://github.com/1600383075/REMD-MWNN/tree/master/data. [ABSTRACT FROM AUTHOR]

This paper introduces three hybrid algorithms that help in solving global optimization problems using reinforcement learning along with metaheuristic methods. Using the algorithms presented, the search agents try to find a global optimum avoiding the local optima trap. Compared to the classical metaheuristic approaches, the proposed algorithms display higher success in finding new areas as well as exhibiting a more balanced performance while in the exploration and exploitation phases. The algorithms employ reinforcement agents to select an environment based on predefined actions and tasks. A reward and penalty system is used by the agents to discover the environment, done dynamically without following a predetermined model or method. The study makes use of Q-Learning method in all three metaheuristic algorithms, so-called RL I−GWO , RL Ex−GWO , and RL WOA algorithms, so as to check and control exploration and exploitation with Q-Table. The Q-Table values guide the search agents of the metaheuristic algorithms to select between the exploration and exploitation phases. A control mechanism is used to get the reward and penalty values for each action. The algorithms presented in this paper are simulated over 30 benchmark functions from CEC 2014, 2015 and the results obtained are compared with well-known metaheuristic and hybrid algorithms (GWO, RL GWO , I-GWO, Ex-GWO, and WOA). The proposed methods have also been applied to the inverse kinematics of the robot arms problem. The results of the used algorithms demonstrate that RL WOA provides better solutions for relevant problems. • Global optimization problems are generally NP-hard problem where metaheuristic algorithms find the optimal solution in the search space. • Reinforcement learning methods give high success rate in finding new global areas compared with metaheuristics and have a more balanced behavior. • Three metaheuristic-reinforcement learning hybrid algorithms are proposed switching between exploration and exploitation phases as and when needed making them more successful in finding better optimized solutions. • They have been applied over 30 benchmark functions and have been simulated to inverse kinematics of the robot arms problem. [ABSTRACT FROM AUTHOR]

Clustering ensemble is an unsupervised ensemble learning method that is very important in machine learning, since it integrates multiple weak base clustering results to produce a strong consistency result. This paper proposes the Markov clustering ensemble (MCE) model to solve the weak stability and robustness of soft clustering ensemble. First, the base clustering algorithms are regarded as new features of the original datasets. Then, the results of the base clustering algorithms are the values of these features, which can break through the framework of consensus cluster ensemble. Second, as the base clustering results are discrete data, the maximum information coefficient is applied to measure their similarity. Accordingly, a graph-based cluster ensemble model can be constructed with row vectors as vertices and the similarity between row vectors as edges. Then, the Markov process can be applied to infer the graph-based cluster ensemble model; therefore, the MCE algorithm is designed according to the inference. To test the performance of the MCE algorithm, clustering algorithms and clustering ensemble algorithms are used to conduct comparative experiments on ten datasets. The experimental results show that the MCE algorithm outperforms the other algorithms in terms of accuracy and purity. [ABSTRACT FROM AUTHOR]

Negative sequential pattern (NSP) discovery is crucial, and sometimes it carries more enlightening information than positive sequential pattern (PSP) mining in data science. Owing to its computational complexity and exponential search space, the task of discovering NSPs is often more difficult and challenging than that for PSPs. To date, a few NSP mining algorithms have been proposed. Particularly, most algorithms only consider a single support for mining, thus they cannot present good results in many special real-world applications. To solve this problem and achieve better efficiency on a long sequence database or a large-scale database, we propose a novel algorithm called Negative Sequential Patterns with Individual Support (NSPIS) in this paper. The projection mechanism is adopted to NSPIS, which allows greatly reduce the search space and simultaneously improve the efficiency. Finally, detailed results of the experiments show that NSPIS can achieve better performance, and it costs less memory on large-scale datasets compared to the state-of-the-art algorithm. [ABSTRACT FROM AUTHOR]

When solving constrained multiobjective optimization problems, it is crucial to deal with several conflicting objectives and various constraints simultaneously. To address these two issues, a constrained multiobjective differential evolution algorithm with an infeasible-proportion control mechanism is presented in this paper. Specifically, two populations are cooperatively employed in the evolution process. The first population is used to solve the original problem, that is, to explore the constrained Pareto front, while the second population is created to search for high-quality objective function information. Furthermore, differential evolution with two specific mutation strategies is employed to update each population. Cooperative strategies can not only balance diversity and convergence but also realize information exchange between the two populations provide new evolutionary directions. In addition, an infeasible-proportion control mechanism is used to gradually decrease the proportion of infeasible solutions in the second population. This enables the main population to pass through the infeasible region barrier in the early evolution stage and facilitates the search for a constrained Pareto front in the later evolution stage. Systematic experiments on 65 benchmark test functions show that the proposed algorithm is superior to or at least comparable to five well-established constrained multiobjective optimization methods. [ABSTRACT FROM AUTHOR]

Over the past decade, the most fundamental Boolean matrix factorization (BMF) algorithms GreCon and GreConD were proposed. Whereas GreConD has become a popular and widely used BMF algorithm, GreCon – the algorithm on which the GreConD is built – is somewhat forgotten in contemporary BMF research; however, GreCon may produce better results than GreConD. The main disadvantage of GreCon algorithm is a slow running time. In the paper, we argue that the search strategy of GreConD, notwithstanding it provides a good result, is limited. We show that the reasons for not using GreCon algorithm are no longer the truth. We revise the algorithm and propose a new approach to storing and updating data required for factor enumeration. By various experiments, we demonstrate that the revised version is competitive with contemporary BMF algorithms in terms of running time. Moreover, in some cases, the revised GreCon outperforms GreConD—one of the fastest BMF algorithms. Furthermore, we show that our novel approach to GreCon enables the utilization of novel approaches to BMF. [ABSTRACT FROM AUTHOR]

In this paper, a general framework of user-based collaborative filtering (CF) is developed with a new p -moment-based similarity measure. The p -moment-based statistics (PMS) of individual rating data are employed to analyze the user rating habit, thereby facilitating the performance improvement of the CF algorithm. On the basis of the PMS, a new yet comprehensive similarity measure is proposed to quantify the distance between two users with focus on both the users' preferences on items and their rating habits. Compared with the traditional ones, our proposed similarity measure is more general with clearer application insights in complicated situations. Furthermore, the weights of different statistics are regarded as adjustable parameters that are determined by utilizing the particle swarm optimization technique so as to achieve good prediction performance. Based on the proposed similarity measure with optimized weights, the neighborhood set consisting of similar users is formed and then the user-based rating prediction is eventually provided. The developed CF algorithm has advantages of high prediction accuracy and wide application potential. Moreover, this CF algorithm is applied on a real-world disease (Friedreich's ataxia) assessment system in order to assist diagnosis for patients with uncertain/missing information. Experimental results demonstrate the validity and efficiency of the proposed algorithm. [ABSTRACT FROM AUTHOR]

HUIM (High utility itemset mining) is a key problem in data mining. The goal is to find itemsets having a high importance or profit in a database, to identify useful knowledge that can support decision-making. In recent years, many HUIM algorithms have been put forward. Among them, utility-list-based algorithms have become very popular as they are easily extendable and efficient. Although several improvements were made, efficiency remains a critical issue. To address this problem, this paper proposes to improve the utility-list construction process, a key operation that has not been much studied in prior work. A novel set of bitwise operations is proposed called BEO (Bit mErge cOnstruction) to speed up the construction process. Besides, a novel data structure called UBP (Utility Bit Partition) is designed to support BEO. This structure is integrated into a novel UBP-Miner algorithm, which also applies several search space reduction strategies. Experimental results show that UBP-Miner is faster than several state-of-the-art algorithms such as HUI-Miner* and ULB-Miner on common benchmark datasets. [ABSTRACT FROM AUTHOR]

The weighted sum coloring problem (WSCP) is an extension of the weighted vertex coloring problem and sum coloring problem, and has important applications in the batch scheduling of distributed systems. Compared to the weighted vertex coloring problem and sum coloring problem, where extensive methods have been proposed for efficiently solving graphs in the real world, fewer solvers have been developed for WSCP, especially no practical complete algorithm is available for solving WSCP. This paper is devoted to design branch-and-bound strategies for solving WSCP, including a branching heuristic rule, a lower bound based on weighted clique partitioning, and a branching search framework for searching all possible coloring solutions. As a result, a new branch-and-bound algorithm BABWSCP is proposed. We also design a novel combination method for improving the local search using complete algorithm in the restart process for WSCP. Consequently, a new local search algorithm named BABLS is designed. Experimental results show that our branch-and-bound strategies are effective, and BABLS almost always outperforms the state-of-the-art WSCP solvers for the benchmarks studied in previous works. • Novel branching heuristics and lower bound based on weighted clique partitioning for weighted sum coloring problem. • An efficient branch-and-bound algorithm BABWSCP for weighted sum coloring problem. • A paradigm of improving local search using branch-and-bound for weighted sum coloring problem. • An algorithm BABLS by improving the local search algorithm CCLSWSC using the branch-and-bound algorithm BABWSCP. [ABSTRACT FROM AUTHOR]

Resampling is the widely used method for imbalance learning. Most existing resampling methods use various techniques in the original sample space to rebalance imbalanced datasets, but they may cause loss of valuable information or aggravate interclass overlap. In this paper, sample concatenation, i.e., concatenating two samples with the same labels into one sample, is introduced into imbalance learning, and a resampling algorithm based on sample concatenation (Re-SC) is proposed. Re-SC transforms an imbalanced training dataset in the original sample space into a concatenated dataset in a new sample space. In the transformation process, Re-SC considers both the distribution of the original dataset and that of the majority samples, thereby alleviating the loss of valuable samples and reducing the class overlapping region. Furthermore, an ensemble resampling algorithm based on sample concatenation (EnRe-SC) for imbalanced data is also proposed. EnRe-SC can reduce the negative effect of removing part of the samples from the majority class. Experiments were conducted on UCI and KEEL imbalanced datasets to evaluate the performance of the proposed methods. The results verify the effectiveness of the proposed methods. • Sample concatenation is introduced and analyzed for imbalance learning. • Two resampling algorithms based on sample concatenation are proposed. • Data complexity is used to measure the classification difficulty of imbalanced data. • A superior performance of the proposed algorithms is verified via experiments. [ABSTRACT FROM AUTHOR]

Today, many web application service providers rely on clouds to deploy applications to serve users. Generally, request arrivals faced by web applications are dynamic and uncertain. When a service provider deploys web applications in clouds, for saving costs, it needs to flexibly rent cloud VM (Virtual Machine) instances based on dynamic request arrivals. However, renting an instance too early may incur more rental fees for the new instance being added incorrectly due to few future requests, and renting an instance too late may incur more penalty fees for SLA (service-level agreement) violations due to too many future requests, which indicates that an arbitrary instance scaling decision will incur more costs. For making optimal instance scaling decisions, future request arrival rate curves are needed, but it is generally very hard to predict them precisely. To solve this problem, in this paper, we propose a cost-driven online auto-scaling algorithm which can make optimized instance rental decisions without requiring future knowledge. We show theoretically that the proposed algorithm can achieve a guaranteed competitive ratio which is less than 2. Eventually, we verify the effectiveness of our online auto-scaling algorithm via extensive experiments using workload data which can simulate real end users. [ABSTRACT FROM AUTHOR]

Image segmentation has been an important technique in the field of image processing. Fine-level manual annotations are very limited and difficult for a large collection of imbalanced images, where each image contains quite different small objects. However, we find that these imbalanced images have similar decision graphs obtained by a lightweight and simple clustering algorithm Density Peaks (DPeaks). Hence, in this paper, we propose a weakly supervised image segmentation algorithm. It trains a decision curve from decision graphs of a few sample imbalanced images by SVM and Support Vector Regression (SVR), which is effective for identifying the sparse region of an imbalanced image. Besides, RangeTree is applied to accelerate RDP for large images due to the high complexity of DPeaks Clustering. Experiments prove that the proposed algorithm works well on imbalanced image data sets, not only it can recognize main things, but also has the ability to identify some relatively small objects. [ABSTRACT FROM AUTHOR]

Surrogate-assisted evolutionary algorithms, which combine the powerful searching ability of evolutionary algorithms (EAs) with the predictive ability of surrogate models, are effective to solve expensive optimization problems. In this paper, an efficient data-driven EA based on multi-evolutionary sampling strategy (DDEA-MESS) is proposed. In DDEA-MESS, three sampling strategies (surrogate-assisted global search strategy, surrogate local search strategy and trust region search strategy) are combined together. The first strategy is focused on global exploration and the other two strategies are focused on local exploitation. Moreover, a probabilistic regulating mechanism for multi-strategy selection is proposed to avoid trapping in local optimum. Finally, a method of update the database is adapted to help algorithm escape from local optimum. The proposed algorithm is compared with other state-of-the-art algorithms on benchmark problems from 30 to 100 dimensions. The result indicate that DDEA-MESS has better performance for solving expensive optimization problems. • A multi-evolutionary sampling strategies algorithm is proposed. • The database update strategy is presented to escape from local optima. • Probabilistic regulating mechanism is proposed for multi-strategy selection. [ABSTRACT FROM AUTHOR]

As an effective tool for the representation and reasoning of uncertain theories, Bayesian networks are widely used in various fields of artificial intelligence. However, learning the optimal Bayesian network structure is an NP-hard problem. The score-and-search approach is a common method for solving the problems associated with Bayesian network structure learning. This paper presents a novel method for learning the structure of a Bayesian network using a discrete firefly optimization algorithm, which has been successfully applied to solve various optimization problems. In the proposed algorithm, each firefly moves in a discrete space based on a redefined movement strategy. Then, the mutation operator is employed to prevent the algorithm from stopping prematurely and falling into a local optimum. Finally, a local optimizer is used to enhance the exploitation ability of the firefly to obtain the best feasible solution. We compared the proposed algorithm with state-of-the-art algorithms on well-known benchmark networks. The experimental results show that the proposed algorithm has better convergence accuracy and higher scores in most cases, compared to other algorithms, indicating that the proposed algorithm can be used as an effective and feasible method for learning Bayesian network structures. [ABSTRACT FROM AUTHOR]

In recent years, green manufacturing has attracted wide attention from researchers. However, the energy efficiency problem in matrix manufacturing workshops is still a blank area. This paper considers a novel automatic guided vehicle (AGV) energy-efficient scheduling problem with release time (AGVEESR) to optimize the three objectives of energy consumption, number of AGVs used and customer satisfaction simultaneously. Considering the development of the AGVEESR, we extract problem-specific knowledge, establish a multiobjective mathematical model, and design a hybrid constructive heuristic. Due to the complexity of the problem, we propose an efficient multiobjective greedy algorithm (MOGA) with effective strategies such as new population initialization, greedy operation, and self-adaptive multiple neighbourhood local search. Meanwhile, an ideal-point-based construction operator in the greedy operation phase is presented to lower the computational complexity. Simulation results show that the proposed MOGA has a tremendously superior performance to the five state-of-the-art algorithms in solving the problem considered. [ABSTRACT FROM AUTHOR]

Density-based clustering algorithms have made a large impact on a wide range of application fields application. As more data are available with rising size and various internal organizations, non-parametric unsupervised procedures are becoming ever more important in understanding datasets. In this paper a new clustering algorithm S-DBSCAN 1 1 A sample code is (will be when ready for publication) available at: http://frederic.rosresearch.free.fr/mydata/homepage/. is proposed in the context of knowledge discovery. S-DBSCAN belongs to the connectivity-based family such as DBSCAN but with noticeable differences and advantages as working in a differential mode. It is formalized via a very simple hierarchical process that hybridizes distance, k -nearest and Density peaks concepts. It aims at partitioning existing data into clusters until no more clustering can be done. The information delivered allows the user to intuitively deduce different sets of natural partitions in clusters at different scales. S-DBSCAN scans the database in a ordered way by applying its algorithm core (S-DBSCANCORE) with judicious input parameters. Given a set of data patterns in some space, S-DBSCANCORE groups together data patterns that are closely packed together with respect to the differential density. Data patterns whose nearest neighbors have too different densities are detected and marked as borders while the others are not visited. S-DBSCAN embeds some intelligence that makes it self-tuning (almost fully automatic) and not dependent on a global density threshold as many existing algorithms. Tests were carried out using 2-dimensional benchmark datasets of various shapes and densities. They showed that S-DBSCAN was highly effective. It also proved efficient in high dimension space when natural clusters exist and much easier to use than competitive algorithms. [ABSTRACT FROM AUTHOR]

The optimal combination of power distribution feeder reconfiguration (PDFR) with distributed generators (DGs) is one of the most attractive combinatorial optimization issues. This paper proposes an improved Heap-based algorithm to enhance the performance of a recently published technique called Heap-based optimizer (HO). HO's performance is enhanced with a Deeper Exploitative Improvement (DEI) by applying an effective exploitation feature to boost the searching around the leader position with the goal of enhancing its global search capabilities and avoiding becoming trapped in a local optimum. The effectiveness of the proposed HODEI in comparison with the conventional HO algorithm is checked on 35 benchmark optimizing functions which have either unimodal and multimodal characteristics. The proposed HODEI and the conventional HO are employed for simultaneous DG allocations and PDFR using two IEEE standard distribution networks of 33 and 69-bus at various loading conditions. Additionally, a large-scale 137-bus distribution network is utilized to assess the validity and efficiency of the proposed HODEI with several independent variables. The proposed HODEI is contrasted with several recent algorithms of harmony search, genetic, equilibrium, marine predators, fireworks, and firefly optimizers. The simulation results demonstrates that the proposed HODEI always gives better performance compared to the conventional HO algorithm. The proposed HODEI shows significant improvement for the voltages at all buses. Furthermore, the proposed HODEI outperforms the compared algorithms in acquiring the minimum best, mean, worst and standard deviations of the considered fitness. [ABSTRACT FROM AUTHOR]

The clustered minimum routing cost tree (CluMRCT) problem is a recent problem with a wide range of real-life applications, especially in designing computer networks with peer-to-peer architecture. Many multifactorial evolutionary algorithms have been proposed to solve multiple CluMRCT problems simultaneously. However, these algorithms only function effectively on complete graphs with small-and-medium sizes. Moreover, the blindness and randomness in the transfer of genetic materials cause a reduction in the exploitation ability and make these algorithms ineffective to solve low-similarity tasks. This paper proposes a hybrid multitasking algorithm named multifactorial firefly algorithm, which integrates the firefly algorithm's strong exploitation ability to enhance the self-evolution of each task when facing low-similarity tasks while improving inter-task knowledge transfers by delivering higher-quality solutions. Also, the proposed algorithm is equipped with new encoding and decoding to focus more on potential search areas on both complete and sparse graphs. The experiments and Wilcoxon signed-rank tests were conducted on various instances to verify our proposal with several state-of-the-art methods. The results portrayed that the proposed encoding scheme helped multitasking algorithms improve solution quality by 32% on average. Besides, the statistical test values proved the superiority of the proposed hybrid algorithm in terms of solution quality and convergence trend. • Develop a new encoding and decoding scheme for the CluMRCT problem, allowing evolutionary algorithms to function and focus on more potential search areas on complete and sparse graphs. • Propose a combination of the multifactorial evolutionary algorithm and firefly algorithm, which enhances the exploitation ability and the quality of knowledge transfers of the evolutionary multitasking algorithm when solving low-similarity tasks. • Evaluate the efficiency of the proposed algorithm and encoding method on various instances. The results proved that the proposed method outperformed all existing approaches in terms of solution quality and convergence trend. [ABSTRACT FROM AUTHOR]

This paper investigates a decentralized supply chain that is composed of one manufacturer and multiple distributors. The manufacturer produces goods and wholesales them to multiple distributors and then the distributors sell products to various markets. The entire production period of the manufacturer is divided into several intervals. The decision-making problem of the entire decentralized supply chain is presented as a two-echelon coordination game network, in which each decision-maker can influence decision-making of other levels. A Stackelberg game framework is proposed to coordinate the decision-making process. And then two nonlinear bi-level programming (BLP) models are developed to find the optimal equilibrium decision scheme by switching the leader and follower roles between the manufacturer and the distributors. The models consider the manufacturer's budget constraints in each interval and the market demands are affected by distributors' selling price and advertising strategies. According to the hierarchy and complexity of bi-level programming problem (BLPP), a nested bi-level method based on hybrid state transition algorithm is proposed to address the BLP models, and mapping approximation strategy is utilized to improve computational efficiency. Finally, the numerical experiments are performed to demonstrate the superiority of the proposed method in terms of accuracy and computational efficiency. [ABSTRACT FROM AUTHOR]