In recent years, remote sensing images have been widely used in fields such as meteorology, agriculture, and environmental monitoring. However, due to the complex backgrounds and varying object scales in remote sensing images, current object detection algorithms often suffer from false detections, missed detections, and low detection accuracy. To address these issues, the paper proposes a novel multi-scale feature synergistic and semantic complementary network (MSC-Net). Specifically, in order to improve the model’s multi-scale detection capabilities, we introduce a global multi-scale interaction structure (GMI). GMI first collects multi-level features from the backbone network and then employs multi-scale aggregation units and transformer layers to extract rich multi-scale feature representations from the global information. Finally, a semantic alignment module is used to achieve collaboration between global multi-scale features and local features, facilitating efficient information exchange between different feature levels. To address background noise interference in remote sensing images, we designed a semantic complementary module (SCM). SCM leverages the semantic similarity between adjacent layers to aggregate salient information from both adjacent and current layers, enhancing the feature representation of objects and suppressing the interference of complex backgrounds. Comprehensive experiments on three major public datasets demonstrate the effectiveness of the proposed method.

The rumor detection task aims to identify unofficial and unconfirmed information that is spreading on social media. At any given moment, different users express their opinions, focusing on some propagation events, and the posts they make gradually form a social network that expands as it grows. Over time, nodes and edges form a dynamic graph that presents different states at different moments. However, most existing research focuses more on the text content, social context, propagation mode, etc., and they ignore the factors from many aspects and do not consider the dynamic relationships implied in the propagation development of social media. To analyze these dynamic properties, this article proposes a dynamic network-based multitask rumor detection method called dynamic GCN-based multitask model with temporal event sharing for rumor detection (DGCN-TES). This method can effectively capture the dynamic patterns of relationships in propagation events and change them over time to detect rumors. It is mainly divided into three modules: 1) dynamic-graph convolutional network (GCN) module, which uses dynamic graph neural network to construct the propagation graph of rumor events at different times, which can better capture the dynamic spatial features that change over time; 2) content-long short-term memory (LSTM), which uses the LSTM network as a benchmark model and has been improved to better capture time-series text features over time and for multitask shared interactions; and 3) temporal event sharing layer is the sharing layer, which uses time step as the basic unit of sharing, and realizes the sharing interaction between dynamic structural features and temporal text features between the first two modules. We tested the method on two real-world rumor detection datasets PHEME and WEIBO, and the final results show that the method improved F1-score by more than 2.63% and 3.91% compared to the other best baselines baseline.

Deep learning models are recently applied to detect rumors on social media based on the information in the posts.However, at the early stage of rumor propagation, due to the lack of responses, the performance of these models often degrades.In this paper, we propose a method based on the conditional generative adversarial network, which can generate the responses like data and help the deep learning models in early detection.On two large-scale Sina Weibo datasets, the proposed method is applied on the existing convolution neural network model, the recurrent neural network model, and the recursive neural network model.The results show that the proposed method can significantly improve the performance of the models in the case of zero response, and has performance superiority in a certain early period.

The layout optimization problem is brought from the design of the recoverable satellite, where a set of objects (equipments or devices) are required to be installed on a circular load board.The aim of the problem is to find a layout of the objects with no interference, less unbalance, and less space occupied.Artificial bee colony (ABC) algorithms show good performance in many engineering problems.In this article, based on the analysis of the solution distribution, a problemspecific knowledge based ABC is proposed, which is configured with special initialization and parameter settings.On an open benchmark with ten instances, the proposed ABC is compared with two widely used algorithms.Its performance outperforms the genetic algorithm on all the instances, and outperforms the quasi-human algorithm on nine instances.

Dynamic multiobjective optimisation deals with multiobjective problems whose objective functions, search spaces, or constraints are time-varying during the optimisation process. Due to wide presence in real-world applications, dynamic multiobjective problems (DMOPs) have been increasingly studied in recent years. Whilst most studies concentrated on DMOPs with only two objectives, there is little work on more objectives. This paper presents an empirical investigation of evolutionary algorithms for three-objective dynamic problems. Experimental studies show that all the evolutionary algorithms tested in this paper encounter performance degradedness to some extent. Amongst these algorithms, the multipopulation based change handling mechanism is generally more robust for a larger number of objectives, but has difficulty in deal with time-varying deceptive characteristics.

Many real world optimization problems are dynamic optimization problems (DOPs) whose optima change over time. In this paper, we propose new variants of differential evolution (DE) to solve DOPs. A hybrid method that combines population core based multi-population strategy and prediction strategy and new local search scheme is introduced into DE to enhance its performance for solving DOPs. The population core based multi-population strategy is useful to maintain the diversity of population by using the multi-population and population core concept. The prediction strategy is useful to rapidly adapt to the dynamic environment by using the prediction area. The local search scheme is useful to improve the searching accuracy by suing the new chaotic local search method. Experimental results on the moving peaks benchmark show that the proposed schemes enhance the performance of DE in the dynamic environments.