The efficient global optimization (EGO) algorithm is a kind of Bayesian optimization algorithm that uses the Kriging interpolation model and expectation improvement (EI) criteria as surrogate model and acquisition function, respectively. However, the greediness of EI criteria can lead the EGO algorithm to fall into local optima. Owing to this, revised expectation improvement (REI) criteria are proposed by introducing a balance factor to adjust the exploitation and exploration of EI criteria, and the corresponding algorithm is called the revised efficient global optimization (REGO) algorithm. In order to motivate exploration, and ensure that the computational cost is acceptable, a Latin hypercube based indicator is proposed to denote a balance factor from the viewpoint of sample distribution. Several test functions and an airfoil optimization problem are applied to verify the performance of the REGO algorithm. The results show that the REGO algorithm has acceptable computational cost, a strong ability to find global optima, and good robustness.
17 January 2023 in Structural and Multidisciplinary Optimization
In previous deep learning based infrared small target detection tasks, the problem of detail loss in pooling and downsampling layers was often overlooked in the feature extraction process, and it was not possible to effectively extract global context and utilize the position information in shallow feature maps and semantic information in deep feature maps of infrared small targets. To solve the above problems, first, we designed Dual channel feature fusion net as the backbone network, which removed pooling layers and used advanced Dual channel feature fusion module as the feature extraction module. Then, at the neck of DC-Net, we designed Cross scale feature fusion net suitable for infrared small targets, which performs feature extraction on the features. The information was fully and effectively fused, and finally, we used Sharpening Algorithm for processing to further enhance the small target area. The experimental results on the SIRST and IRIS datasets show that DC Net exhibits superior performance in both qualitative and quantitative analysis. The mAP50 obtained in SIRST testing increased by 11.7%, to 81.4%, on the basis of YOLOv5s. In IRIS, it was 3.7% higher than YOLOv5s, to 96.3%.
29 January 2024 in Structural and Multidisciplinary Optimization
Journal of Simulation (JOS) publishes articles, technical notes, case studies and tutorials in the field of simulation, with a particular focus on the applications of simulation and modelling. Journal of Simulation is an official journal of the OR Society.