查看更多>>摘要：Closed loop layout problem (CLLP) is an important class of design problems encountered in flexible manufacturing system. It is to determine the locations of manufacturing cells along a closed material handling loop. Existing studies on CLLPs typically ignore clearances between adjacent cells and do not optimize the dimension of the closed loop. However, separating adjacent cells with suitable clearances may achieve a better solution with lower material flow cost, and optimizing cell sequence and the dimension of closed loop simultaneously can also improve the solution quality. In this paper, a mixed integer programming formulation (MIP) is established for the CLLP. Then, based on the formulation, a hybrid approach combining an improved differential evolution algorithm and an exact approach (iDE-EA) is proposed to solve the CLLP. The iDE with a hybrid coding is utilized to optimize the dimension of closed loop and the placement sequence of cells simultaneously. The EA is to determine the exact location of each cell (i.e., the clearance between each pair of adjacent cells). iDE-EA is verified on 10 problem instances, and experiments show that iDE-EA improves upon the best-known material handling cost by 0–9.97% compared against the traditional differential evolution algorithm and three state-of-the-art approaches. In addition, using EA to further optimize clearances among cells in the solution obtained by iDE can achieve a more competitive layout in most cases.
查看更多>>摘要：Liver segmentation on images acquired using computed tomography (CT) and magnetic resonance imaging (MRI) plays an important role in clinical management of liver diseases. Compared to MRI, CT images of liver are more abundant and readily available. However, MRI can provide richer quantitative information of the liver compared to CT. Thus, it is desirable to achieve unsupervised domain adaptation for transferring the learned knowledge from the source domain containing labeled CT images to the target domain containing unlabeled MR images. In this work, we report a novel unsupervised domain adaptation framework for cross-modality liver segmentation via joint adversarial learning and self-learning. We propose joint semantic-aware and shape-entropy-aware adversarial learning with post-situ identification manner to implicitly align the distribution of task-related features extracted from the target domain with those from the source domain. In proposed framework, a network is trained with the above two adversarial losses in an unsupervised manner, and then a mean completer of pseudo-label generation is employed to produce pseudo-labels to train the next network (desired model). Additionally, semantic-aware adversarial learning and two self-learning methods, including pixel-adaptive mask refinement and student-to-partner learning, are proposed to train the desired model. To improve the robustness of the desired model, a low-signal augmentation function is proposed to transform MRI images as the input of the desired model to handle hard samples. Using the public datasets, our experiments demonstrated the proposed unsupervised domain adaptation framework reached four supervised learning methods with a Dice score 0.912 ± 0.037 (mean ± standard deviation).
查看更多>>摘要：Particle swarm optimization (PSO) has shown its advantages in various optimization problems. Topology and updating strategies are among its key concepts and have significant impacts on optimization ability. This paper proposes a pyramid PSO (PPSO) with novel competitive and cooperative strategies to update particles’ information. PPSO builds a pyramid and assigns each particle to a specific layer according to its fitness. The particles at the same layer will make a pairwise comparison to determine the winners and the losers. The losers will cooperate with their corresponding winners, while the winners will cooperate with the particles at the upper layer and those at the top layer. Each particle in PPSO has its own learning behavior, having more than one exemplar rather than the only global best to learn from. The diversity of the swarm is enhanced and it positively affects the performance of PSO. Extensive experiments demonstrate that the PPSO has superior performance in terms of accuracy, Wilcoxon signed-rank test and convergence speed, yet achieves comparable running time in most cases, when compared with the canonical PSO and eight state-of-the-art PSO variants. Furthermore, we analyze the influence of parameters for the PPSO. All these illustrate that the PPSO is promising for numerical optimization.
查看更多>>摘要：State transition algorithm (STA) is an efficient and powerful metaheuristic method for solving global optimization problems, and it has been successfully applied in many engineering fields in the past few years. However, the basic STA has weak local search capability and shows slow convergence rate and low convergence accuracy in the later search stage. In view of the above shortcomings, an adaptive state transition algorithm (ASTA) with local enhancement is proposed in this paper. Firstly, the order of using state transformation operators and the optimal parameters of the operators are considered in each iteration of ASTA, and a statistical method is employed to adaptively select the optimal transformation operator and the parameter values of the optimal operator to speed up the search process. Then, an adaptive call strategy is adopted to determine its convergence to the neighborhood of the optimal solution and to decide whether to perform the quasi-Newton operator for local enhancement. Finally, the degree to which the current solution is close to the optimal solution is judged by the information of historical solutions, and an analytical solution is quickly obtained by calling the quadratic interpolation operator. The effectiveness of the proposed ASTA is checked, through a comparison with other metaheuristic methods, on 15 benchmark functions and several real-world optimization problems. Experimental results show that ASTA has a stronger search capability than the basic STA, STA variants, and some state-of-the-art metaheuristic methods.
查看更多>>摘要：The evolutionary algorithms (EAs) have been shown favorable performance for feature selection. However, a large number of evaluations are required through the EAs. Thus, they will be inappropriate to optimize feature selection when the size of data set is large. In this paper, we propose a multi-surrogate assisted binary particle swarm optimization, denoted as MS-assisted DBPSO. Two surrogate models are trained, which are utilized to approximate the fitness values of the individuals in two sub-populations, respectively. After that, a new population will be generated by the communication between the two sub-populations. Furthermore, dynamic transfer function is proposed in this paper to balance global and local search aiming to find optimal solution with limited computational resource. The experimental results on binary benchmark functions and the feature selection in the UCI data sets demonstrate that our proposed method is efficient on reducing running time and prediction error.
查看更多>>摘要：Good visual perception capability for object plays an important role in binary segmentation task, such as the segmentation for portraits and pulmonary nodules. When facing the same object in different backgrounds, humans always keep consistent visual perception. This has motivated semantic data augmentation strategies widely used in segmentation tasks. For example, ‘Cut-Paste’ strategy creates many images by changing background and assigns them to the same segmentation ground-truths for enhancing training. However, even if using these strategies, there are still differences among the segmentation results of images with the same object and different backgrounds. Hence, this paper proposes to adopt image-level classification and visual attention consistency under background-change to enhance the training of binary segmentation. The combination of image-level classification and class activation mapping can activate and visualize certain regions, which are related to classification label. The visual attention consistency requires the activated object attention to keep consistent when background of the input image changes. Based on this purpose, we augment the dataset by changing backgrounds with ‘Cut-Paste’. Afterwards, we adopt a shared triple-branch network to make original image, background-cut-out image and background image as inputs, and then propose image-level classification and attention consistency to train the binary segmentation network. Experimental results based on two datasets demonstrate that our method achieves new state-of-the-art binary segmentation performance.
查看更多>>摘要：Earthquakes are random triggering phenomenons that generate clusters in space and time, thus creating a bias in a seismic catalog. Seismic de-clustering separates seismic catalog into mainshocks, aftershocks–foreshocks, and backgrounds, widely used in earthquake prediction models and seismic hazard assessment. The segregation of an optimal number of earthquake clusters and backgrounds is formulated as an unsupervised problem. This manuscript introduces a multi-objective chimp-optimization algorithm (MOCOA) to de-cluster seismicity of earthquake-prone regions. The chimp optimization is inspired by the natural hunting behavior of the chimp to catch the prey. The algorithm effectively balances the exploration of search space (Driving and Chasing) and exploitation around the best solution achieved so far (Attack). In MOCOA, archive controller and archive grid-based approaches are incorporated for selecting non-dominated solutions. The proposed MOCOA is tested on fifteen mathematical test problems and compared with popular algorithms like MOEA/D, MODE, MOPSO, and SPEA2. The binary version of MOCOA is designed for the de-clustering problem. In the time and space domain, respectively, two objective functions, m-Morisita Index (m-MI) and coefficient of variance (COV), are analyzed. The proposed de-clustering algorithm is applied on thirty-two-year historical seismic catalogs of the Himalayas, California, Indonesia, Japan, Iran, and Mexico. Comparative analysis between five existing benchmark de-clustering techniques is performed to check the potential of the proposed MOCOA. The simulation results generated by the proposed algorithm show that obtained de-clustered catalogs COV values lying near unity and m-MI values achieved maximum values. Validation of results using cumulative plots, lambda plots, and inter-event time versus inter-event distance plots signify the accurate discrimination of aftershocks and background events in the catalogs.
查看更多>>摘要：Ransomware is malware that encrypts the victim's data and demands a ransom for a decryption key. The increasing number of ransomware families and their variants renders the existing signature-based anti-ransomware techniques useless; thus, behavior-based detection techniques have gained popularity. A difficulty in behavior-based ransomware detection is that hundreds of thousands of system calls are obtained as analysis output, making the manual investigation and selection of ransomware-specific features infeasible. Moreover, manual investigation of the analysis output requires domain experts, who are expensive to hire and unavailable in some cases. Machine learning methods have shown success in a wide range of scientific domains to automate and address the problem of feature selection and extraction from noisy and high-dimensional data. However, automated feature selection is under-explored in malware detection. This study proposes an automated feature selection method that utilizes particle swarm optimization for behavior-based ransomware detection and classification. The proposed method considers the significance of various feature groups of the data in ransomware detection and classification and performs feature selection based on groups’ significance. The experimental results show that, in most cases, the proposed method achieves comparable or significantly better performance than other state-of-the-art methods used in this study for benchmarking. In addition, this article presents an in-depth analysis of the significance of various features groups and the features selected by the proposed method in ransomware detection and classification.
查看更多>>摘要：This work addresses a novel General Variable Neighborhood Search (GVNS) solution method, which integrates intelligent adaptive mechanisms to re-order the search operators during the intensification and diversification phases, in an effort to enhance its overall efficiency. To evaluate the performance of the new GVNS scheme, asymmetric and symmetric instances of the classic Traveling Salesman Problem (TSP) from the TSPLib were solved. The obtained results of the Double-Adaptive GVNS were compared with those achieved by two single-adaptive GVNS, which use an adaptive mechanism either for the intensification or the diversification phase and with a conventional GVNS. For a fair comparison, all GVNS schemes were structured using the same local search and shaking operators. Moreover, the novel GVNS algorithm was compared with some recent solution methods for the TSP, found in the open literature. The comparative studies revealed the high efficiency of the novel VNS scheme and underlined the significant impact of intelligent mechanisms on the performance of classic metaheuristic frameworks.
查看更多>>摘要：Parallel computing problems of multiphysical coupling applications based on discrete grids can be equivalently transformed into parallel computing problems based on a directed acyclic graph (DAG). Due to the problem of the discrete high-dimensional grids of the multiphysical coupling application, the directed data dependencies usually have parallelism. Moreover, this kind of scheduling problem based on DAG is NP-hard problem. Heuristic algorithms are often used to achieve optimal execution order scheduling of tasks and mapping between tasks and processors. In this paper, we propose an improved chemical reaction optimization algorithm based on adaptive search strategy (ASSCRO), which is used to solve the DAG task scheduling problem of discrete multiphysical coupling applications. ASSCRO is divided into two phases. The first phase is used to search the directed execution order of the tasks, and the second phase aims to use the heuristic strategy to map the tasks to the processors efficiently. In the four basic reactions of CRO, the algorithm can cover a larger search solution space by using an adaptive search strategy, it can obtain a better solution, and achieve less overhead and superior performance than the state-of-the-art. We conducted the experiment that applied our ASSCRO to deal with the multiphysical coupling applications. The experimental results showed that the proposed algorithm outperforms other algorithms in multiple metrics when dealing with DAG scheduling problems.
NSTL
Elsevier