Recently, deep learning has achieved remarkable breakthroughs in hyperspectral image (HSI) classification tasks, particularly with methods based on convolutional neural networks (CNNs) and transformers. However, these methods have several limitations: 1) the limited receptive field inherent in the convolutional layer greatly hampers capturing feature contextual information on a large scale and 2) transformers cannot establish strong local relationships, making it challenging to characterize complex dependencies between distant pixels and different bands in HSIs. Moreover, as the network complexity increases, so does the number of network parameters. We propose a novel network called the spectral-spatial convolutional fusion long-range dependence transformer network (LRDTN) for HSI classification to address these challenges. LRDTN comprises three key components: dynamic-dependent convolutional (DDC) module, the multiscale enhanced fusion (MsEF) module, and the local–global perception transformer (LGPT). Specifically, the DDC dynamically models local features, while the MsEF integrates information from different scales to capture contextual relationships in HSI features effectively. Additionally, the ability to mine and utilize HSI local-global features and complex long-range dependencies is enhanced by the proposed transformer variant, LGPT. Ultimately, through the ingeniously designed structure of the LRDTN, the model effectively maintains its performance while reducing the number of network parameters. Extensive experiments conducted on four typical HSI datasets, including urban areas, agricultural areas, and swamps, demonstrate the superiority of LRDTN over other state-of-the-art networks. The code is available athttps://github.com/ybyangjing/LRDTN.

Compositional Zero-Shot Learning (CZSL) has been applied to various scenarios, including scene understanding, visual-language representation, and domain adaptation. Despite numerous endeavours and significant advancements, the crucial issues of fuzzy conceptualization of visual attributes and insufficient inter-class connectivity, have remained insufficiently addressed. To address these issues, we propose Learning Visual Attributes Representation for Compositional Zero-Shot Learning (LVAR-CZSL), which has the ability to learn visual attributes and inter-class dependencies. LVAR-CZSL is mainly composed of two key components: the Visual Attribute Representation Module (VARM) and the Connected Learning Module (CLM). Specifically, VARM extracts detailed attributes and object visual features from global visual features, resolving the issue of fuzzy visual attribute concepts. Moreover, CLM endows LVAR-CZSL with the capability to perceive connectivity between different attributes and objects, effectively enhancing inter-class connectivity. To establish a close connection between VARM and CLM and minimize the gap between image and text features, we introduce the composition-attribute-object Joint Scoring Function (JSF). Additionally, we propose Joint Loss Function (JLF) to optimize the learning process of VARM and CLM. The experiment results on four datasets show that LVAR-CZSL achieves state-of-the-art performance. The code is available athttps://github.com/mxjmxj1/LVAR-CZSL.

Abstract Few‐shot semantic segmentation aims at recognizing the object regions of unseen categories with only a few annotated examples as supervision. The key to few‐shot segmentation is to establish a robust semantic relationship between the support and query images and to prevent overfitting. In this paper, an effective multi‐similarity hyperrelation network (MSHNet) is proposed to tackle the few‐shot semantic segmentation problem. In MSHNet, a new generative prototype similarity (GPS) is proposed, which, together with cosine similarity, establishes a strong semantic relationship between supported images and query images. In addition, a symmetric merging block (SMB) in MSHNet is proposed to efficiently merge multi‐layer, multi‐shot, multi‐similarity features to generate hyperrelation features for semantic segmentation. Experimenting on two benchmark semantic segmentation datasets (Pascal − 5 i and COCO − 20 i ) shows that this method achieves a mean intersection‐over‐union score of 72.3% and 56.0%, respectively, which outperforms the state‐of‐the‐art methods by 1.9% and 6.5%.

The accurate prediction of traffic flow is paramount for the advancement of intelligent transportation systems. Despite this, current prediction models only account for either temporal or spatial features in isolation, without considering their interaction, impeding the model’s ability to express itself. In light of this, we propose the graph differential equations network (GDENet), an approach that can effectively mine spatiotemporal correlation. Specifically, we propose a spatiotemporal feature integrator (STFI), which alleviates the error caused by the deviation of the sampling distribution from the overall distribution. By incorporating temporal information into the model for training and combining it with spatial features, we thoroughly explore the spatiotemporal intrinsic association. When compared to state‐of‐the‐art methods, our proposed algorithm reduces memory consumption and elevates computational efficiency and the practical value. We conduct experiments with real‐world datasets, and our proposed model outperformed advanced prediction models.

Most existing DTN routing algorithms can't show efficient network performance. Prophet is a routing protocol widely used for DTNs. The delivery predictability of active nodes will obviously reduce with the same speed as less active nodes. In this paper, we propose an Adaptive Probability Prediction Routing (APPR) approach, which can dynamically adjust the aging factor based on the active degree of nodes. Leveraging the APPR approach, the delivery predictability of each node can be reduced in different speed if the node does not encounter the other node. Furthermore, when two nodes encounter, the forward decision depends on the distance between each node to the destination node, if the delivery predictabilities of encountered nodes are similar. Compared with the traditional Prophet, simulation results demonstrate the proposed APPR approach can significantly improve successful delivery ratio, reduce overhead ratio and the average latency.

In certain applications, classification models have to be trained with small datasets. This study proposes a new deep neural network with a feature generalisation layer (FGL). First, instead of using a generative network for data augmentation, the FGL is modelled using a latent variable model to diversify features directly by sharing other layers. Then, dual‐objective functions are defined to optimise the parameters of the network: one minimises the generation error and the other minimises the classification error. Finally, a parallel multibranch structure is used in the FGL to improve the convergence of model training. The classification accuracy obtained using various quantities of training samples increased up to 4.63% on the MNIST dataset, up to 3.00% on the CIFAR10 nature image dataset, over the reference model. These experimental results illustrate the effectiveness of the authors’ method for training classification models with small datasets.

The combination of linear range walk correction and keystone transform is a good choice to focus high-resolution highly squint synthetic aperture radar (SAR) data because it is an effective way to remove linear range cell migration (RCM) completely and mitigate range-azimuth coupling. However, the results of this kind of imaging algorithm produce 2-D-variant residual RCM and variant-dependence Doppler phases. To obtain high-quality SAR image, an improved imaging algorithm using an azimuth-variant residual RCM correction (RCMC) and an extended nonlinear chirp scaling (ENLCS) is proposed in this letter. A new circle model is constructed to analyze the azimuth-variant properties of the residual high-order RCM and the Doppler phases. Based on this circle model, an azimuth-variant residual RCMC is implemented by multiplying a fourth-order phase function, and an improved ENLCS is derived to accomplish the azimuth equalization for azimuth compression. Simulation results validate the excellent performance of the proposed algorithm.

An incremental selection hybrid decode-amplify forward (ISHDAF) scheme for the two-hop single relay systems and a relay selection strategy based on the hybrid decode-amplify-and-forward (HDAF) scheme for the multirelay systems are proposed along with an optimized power allocation for the Internet of Thing (IoT). Given total power as the constraint and outage probability as an objective function, the proposed scheme possesses good power efficiency better than the equal power allocation. By the ISHDAF scheme and HDAF relay selection strategy, an optimized power allocation for both the source and relay nodes is obtained, as well as an effective reduction of outage probability. In addition, the optimal relay location for maximizing the gain of the proposed algorithm is also investigated and designed. Simulation results show that, in both single relay and multirelay selection systems, some outage probability gains by the proposed scheme can be obtained. In the comparison of the optimized power allocation scheme with the equal power allocation one, nearly 0.1695 gains are obtained in the ISHDAF single relay network at a total power of 2 dB, and about 0.083 gains are obtained in the HDAF relay selection system with 2 relays at a total power of 2 dB.