Complex & Intelligent Systems最新文献

Thermal layout optimization problems are common in integrated circuit design, where a large number of electronic components are placed on the layout, and a low temperature (i.e., high efficiency) is achieved by optimizing the positions of the electronic components. The operating temperature value of the layout is obtained by measuring the temperature field from the expensive simulation. Based on this, the thermal layout optimization problem can be viewed as an expensive combinatorial optimization problem. In order to reduce the evaluation cost, surrogate models have been widely used to replace the expensive simulations in the optimization process. However, facing the discrete decision space in thermal layout problems, generic surrogate models have large prediction errors, leading to a wrong guidance of the optimization direction. In this work, the layout scheme and its temperature field are represented by images whose relation can be well approximated by a deep neural network. Therefore, we propose an online deep surrogate-assisted optimization algorithm for thermal layout optimization. First, the iterative local search is developed to explore the discrete decision space to generate new layout schemes. Then, we design a deep neural network to build an image-to-image mapping model between the layout and the temperature field as the approximated evaluation. The operating temperature of the layout can be measured by the temperature field predicted by the mapping model. Finally, a segmented fusion model management strategy is proposed to online updates the parameters of the network. The experimental results on three kinds of layout datasets demonstrate the effectiveness of our proposed algorithm, especially when the required computational budget is limited.

Over the last decade, particle swarm optimization has become increasingly sophisticated because well-balanced exploration and exploitation mechanisms have been proposed. The sequential quadratic programming method, which is widely used for real-parameter optimization problems, demonstrates its outstanding local search capability. In this study, two mechanisms are proposed and integrated into particle swarm optimization for single-objective numerical optimization. A novel ratio adaptation scheme is utilized for calculating the proportion of subpopulations and intermittently invoking the sequential quadratic programming for local search start from the best particle to seek a better solution. The novel particle swarm optimization variant was validated on CEC2013, CEC2014, and CEC2017 benchmark functions. The experimental results demonstrate impressive performance compared with the state-of-the-art particle swarm optimization-based algorithms. Furthermore, the results also illustrate the effectiveness of the two mechanisms when cooperating to achieve significant improvement.

Traditional engine cycle innovation is limited by human experiences, imagination, and currently available engine component performance expectations. Thus, the engine cycle innovation process is quite slow for the past 90 years. In this work, we propose a mixed variable multi-objective evolutionary optimization method for automatic engine cycle design. In the first, a simulation toolkit is developed for performance evaluation of potentially viable engine cycle solutions. Then, the engine cycle solutions are mixed encoded by the pins and the parameters of different engine components. The new engine cycle solutions are generated through the mutation operator. Finally, we construct two optimization objectives to drive the optimization process. Through the experimental research, new engine cycle solutions are discovered that exceed the performance of known turbojet and turbofan engines.

Multi-object tracking (MOT) aims to locate and identify objects in videos. As deep learning brings excellent performances to object detection, the tracking-by-detection (TBD) has gradually become a mainstream tracking framework. However, some drawbacks still exist in the current TBD framework: (1) inaccurate prediction of the bounding boxes would occur in the detection part, which is caused by overlooking the actual pedestrian ratio in the surveillance scene. (2) The width of the bounding boxes in the next frame might be indirectly predicted by the aspect ratio, which increases the error of width prediction in the motion prediction part. (3) Association is only performed for high-confidence detection boxes, and the low-confidence boxes caused by occlusion are discarded in the data association part, resulting in fragmentation of trajectories. To address the above issues, we propose a multi-target tracking model incorporating motion estimation and multi-stage association (MEMA). First, the aspect ratio of the ground-true bounding box is introduced to improve the fit of the detection and the ground-true bounding box, and we design the elliptical Gaussian kernel to improve the positioning accuracy of the object center point. Then, the prediction state vector of the Kalman filter is modified to predict the width and its corresponding velocity directly. It can reduce the width error of the prediction box and eliminate the velocity error of the motion estimation, which leads to a more pedestrian-friendly prediction bounding box. Finally, we propose a multi-stage association strategy to correlate different confidence boxes. Without using the appearance feature, the strategy can reduce the impact of occlusion and improve the tracking performance. On the MOT17 test set, the method proposed in this paper achieves a MOTA of 74.3% and an IDF1 of 72.4%, outperforming the current SOTA.

Automated mental retardation (MR) assessment is potential for improving the diagnostic efficiency and objectivity in clinical practice. Based on the researches on abnormal behavior characteristics of patients with MR, we propose an extension and supplement shift multi-scale G3D (ESS MS-G3D) network for video-based assessment of MR. Specifically, all videos are collected from clinical diagnostic scenarios and the skeleton sequence of human body is extracted from videos through an advanced pose estimation model. To solve the shortcomings of existing behavior characteristic learning methods, we present: (1) three G3D styles, enable the network to have different input forms; (2) two G3D graphs and two extension graphs, redefine and extend the graph structure of spatial–temporal nodes; (3) two learnable parameters, realize adaptive adjustment of graph structure; (4) a shift layer, enable the network to learn global features. Finally, we construct a three-branch model ESS MS-STGC, which can capture the discriminative spatial–temporal features and explore the co-occurrence relationship between spatial and temporal domains. Experiments in clinical video data set show that our proposed model has good performance in MR assessment and is superior to the existing vision-based methods. In two-classification task, our model with joint stream achieves the highest accuracy of (94.63%) in validation set and (89.13%) in test set. The results are further improved to (96.52%) and (93.22%), respectively, by utilizing multi-stream fusion strategy. In four-classification task, our model obtains Top1 accuracy of (78.84%) and Top2 accuracy of (91.34%) in test set. The proposed method provides a new idea for clinical mental retardation assessment.

There are inconsistent tasks and insufficient training in the SAR ship detection model, which severely limit the detection performance of the model. Therefore, we propose a twin branch network and design two loss functions: regression reverse convergence loss and classification mutual learning loss. The twin branch network is a simple but effective method containing two components: twin regression network and twin classification network. Aiming at the inconsistencies between training and testing in regression branches, we propose a regression reverse convergence loss (RRC Loss) based on twin regression networks. This loss can make multiple training samples in the twin regression branch converge to the label from the opposite direction. In this way, the test distribution can be closer to the training distribution after processing. For inadequate training in classification branch, Inspired by knowledge distillation, we construct self-knowledge distillation using a twin classification network. Meanwhile, our proposed classification mutual learning loss (CML Loss) enables the twin classification network not only to conduct supervised learning based on the label but also to learn from each other. Experiments on SSDD and HRSID datasets prove that, compared with the original method, the proposed method can improve the AP by 2.7–4.9% based on different backbone networks, and the detection performance is better than other advanced algorithms.

One of the most significant and complete approaches to accommodate greater uncertainty than current fuzzy structures is the T-Spherical Fuzzy Set (T^SPFS). The primary benefit of T^SPFS is that current fuzzy structures are special cases of it. Firstly, some novel T^SPF power Heronian mean (T^SPFPHM) operators are initiated based on Aczel–Alsina operational laws. These aggregation operators (AOs) have the capacity to eliminate the impact of uncomfortable data and can simultaneously consider the relationships between any two input arguments. Secondly, some elementary properties and core cases with respect to parameters are investigated and found that some of the existing AOs are special cases of the newly initiated aggregation operators. Thirdly, based on these AOs and Aczel–Alsina operational laws a newly advanced technique for order of preference by similarity to ideal solution (TOPSIS)-based method for dealing with multi-attribute group decision-making (MAGDM) problems in a T-Spherical fuzzy framework is established, where the weights of both the decision makers (DMs) and the criteria are completely unknowable. Finally, an illustrative example is provided to evaluate and choose the pharmaceutical firms with the capacity for high-quality, sustainable development in the T^SPF environment to demonstrate the usefulness and efficacy. After that, the comparison analysis with other techniques is utilized to demonstrate the coherence and superiority of the recommended approach.

In emergency management, the transportation scheduling of emergency supplies and relief personnel can be regarded as the multi-objective shortest path problem with mixed time window (MOSPPMTW), which has high requirements for timeliness and effectiveness, but the current solution algorithms cannot simultaneously take into account the solution accuracy and computational speed, which is very unfavorable for emergency path decision-making. In this paper, we establish MOSPPMTW matching emergency rescue scenarios, which simultaneously enables the supplies and rescuers to arrive at the emergency scene as soon as possible in the shortest time and at the smallest cost. To solve the complete Pareto optimal surface, we present a ripple spreading algorithm (RSA), which determines the complete Pareto frontier by performing a ripple relay race to obtain the set of Pareto optimal path solutions. The proposed RSA algorithm does not require an initial solution and iterative iterations and only needs to be run once to obtain the solution set. Furthermore, we prove the optimality and time complexity of RSA and conduct multiple sets of example simulation experiments. Compared with other algorithms, RSA performs better in terms of computational speed and solution quality. The advantage is especially more obvious in the computation of large-scale problems. It is applicable to various emergency disaster relief scenarios and can meet the requirements of fast response and timeliness.

This article presents model-free adaptive control based on an intuitionistic fuzzy neural network for nonlinear systems with event-triggered output. Essentially, model-free adaptive control (MFAC) is constructed by establishing an online approximate model of the controlled system using the pseudo-partial derivative (PPD) form. By the proposed scheme, first, an intuitionistic fuzzy neural network (IFNN) is developed as an estimator for time-varying PPD in both compact-form dynamic linearization (CFDL) and partial-form dynamic linearization (PFDL) for the MFAC technique. Second, two periodic event-triggered output methods are integrated with the proposed IFNN-based MFAC in both forms to save communication resources and reduce the computation burden and energy consumption. Based on the Lyapunov theory and BIBO stability approach, necessary conditions are established to guarantee the convergence of the adaptive law of the IFNN controller and the boundary of the tracking error of the closed loop system. Third, regarding the feasibility and the effectiveness of the developed control method, two simulation examples including the continuous stirred-tank reactor (CSTR) system and the heat exchanger system are given. Finally, the practical validation of the proposed data-driven control method is conducted via the speed control of a DC motor.

Recent improvements in networking technologies have led to a significant shift towards distributed cloud-based services. However, adequate management of computation resources by providers is vital to maintain the costs of operations and quality of services. A robust system is needed to forecast demand and prevent excessive resource allocations. Extensive literature review suggests that the potential of recurrent neural networks with attention mechanisms is not sufficiently explored and applied to cloud computing. To address this gap, this work proposes a methodology for forecasting load of cloud resources based on recurrent neural networks with and without attention layers. Utilized deep learning models are further optimized through hyperparameter tuning using a modified particle swarm optimization metaheuristic, which is also introduced in this work. To help models deal with complex non-stationary data sequences, the variational mode decomposition for decomposing complex series has also been utilized. The performance of this approach is compared to several state-of-the-art algorithms on a real-world cloud-load dataset. Captured performance metrics ((R^2), mean square error, root mean square error, and index of agreement) strongly indicate that the proposed method has great potential for accurately forecasting cloud load. Further, models optimized by the introduced metaheuristic outperformed competing approaches, which was confirmed by conducted statistical validation. In addition, the best-performing forecasting model has been subjected to SHapley Additive exPlanations analysis to determine the impact each feature has on model forecasts, which could potentially be a very useful tool for cloud providers when making decisions.