Few-shot learning (FSL) aims to recognize novel concepts from only a few labeled support samples. Recent studies enhance support features by incorporating additional semantic information (e.g., class descriptions) or designing complex semantic fusion modules. However, these methods still suffer from hallucinating semantics that contradict the visual evidence due to the lack of grounding in actual instances, resulting in noisy guidance and costly corrections. To address these issues, we propose a novel framework, bridging Vision and Text with LLMs for Few-Shot Learning (VT-FSL), which constructs precise cross-modal prompts conditioned on Large Language Models (LLMs) and support images, seamlessly integrating them through a geometry-aware alignment mechanism. It mainly consists of Cross-modal Iterative Prompting (CIP) and Cross-modal Geometric Alignment (CGA). Specifically, the CIP conditions an LLM on both class names and support images to generate precise class descriptions iteratively in a single structured reasoning pass. These descriptions not only enrich the semantic understanding of novel classes but also enable the zero-shot synthesis of semantically consistent images. The descriptions and synthetic images act respectively as complementary textual and visual prompts, providing high-level class semantics and low-level intra-class diversity to compensate for limited support data. Furthermore, the CGA jointly aligns the fused textual, support, and synthetic visual representations by minimizing the kernelized volume of the 3-dimensional parallelotope they span. It captures global and nonlinear relationships among all representations, enabling structured and consistent multimodal integration. The proposed VT-FSL method establishes new state-of-the-art performance across ten diverse benchmarks, including standard, cross-domain, and fine-grained few-shot learning scenarios. Code is available at https://github.com/peacelwh/VT-FSL.

To promote consensus, various consensus feedback models driven by social relationships (SRs) have been proposed for social network (SN) group decision-making (SNGDM). However, insufficient attention has been paid to how decision-makers (DMs) interact and to the utility they derive from such social interactions. Therefore, a consensus decision process considering social interaction probability and individual utility is proposed. First, this article explores the direct and indirect social interaction patterns among DMs. A quantum-like Bayesian network (BN) is constructed within the SN environment and combined with similarity effect to infer the probabilities of social interactions. Second, individual utility functions are developed based on two dimensions: 1) decision outcomes and 2) social interaction gains. Third, a two-stage consensus feedback model is proposed. It matches reference opinions by assessing the necessity of establishing new SRs and uses adaptive feedback coefficients to balance group consensus with individual utility. Finally, an illustrative example verifies the feasibility of the proposed model. Subsequent comparisons and simulations further demonstrate its effectiveness and advantages.

Image registration plays a pivotal role in various image processing applications, which is widely used in image fusion and change detection. However, The presence of speckle noise in SAR images causes a primary reduction in registration accuracy and existing algorithms have not achieved high-precision registration while maintaining low computational complexity. This paper proposes an image registration algorithm based on edge intersections and deep learning descriptors. An edge-directed voting mechanism is introduced to identify corner points, and a custom SAR image dataset is constructed to retrain the HardNet descriptor network. Experimental results validate the superiority of the proposed method in terms of robustness and accuracy, achieving SAR image registration on a self-constructed dataset with an RMSE of 0.38, showcasing the utmost registration accuracy, while maintaining lower computational complexity than traditional approaches.

Crowdshipping problem can be challenging as the platform are continuously but sporadically receiving crowdshippers and delivery tasks with heterogeneous origin and destination. In this paper, the dynamic multi-vehicle pickup and delivery problem with crowdshippers (DMV-PDPC) is considered. Leveraging the deep reinforcement learning framework, the attention model with centralized vehicle network (AMCVN) method is developed. Unlike traditional heuristic or existing vehicle-changing methods, AMCVN integrates a centralized vehicle network (CVN) that can observe the state information of all vehicles, enhancing its overall performance. In each decision-making step, the CVN monitors the state of the vehicles and selects one of the vehicles. Subsequently, the attention-based route generating network (RGN) determines the next node to be visited by the chosen vehicle. Instead of using a penalty term in the reward function to regulate the sequence of visits to pickup and delivery nodes, a more precise control method, namely the rolling mask scheme (RMS), is implemented. The method’s evaluation is carried out via a simulation experiment using a real-world road network. This evaluation demonstrates that the proposed method effectively tackles the DMV-PDPC challenge, outperforming current state-of-the-art learning-based models and heuristic methods. Moreover, the method shows exceptional generalization capabilities, as evidenced by its adaptability to different numbers of tasks and vehicles.

Most existing opinion formation processes apply one opinion dynamics model. However, this article combines opinion formation and complex networks to innovatively develop two new opinion dynamics models to more realistically describe the opinion evolution process: 1) an opinion similarity mixed (OSM) model and 2) a structural similarity mixed (SSM) model, both of which include characteristics from the DeGroot model and the Hegselmann–Krause bounded confidence model. In addition, the strong and weak relations between individuals are considered. The network dynamically changes by two developed network updating algorithms based on opinion similarity and structural similarity. Simulations are then conducted using artificial and real-world networks, which are Erdös-Rényi random networks, random regular networks, scale-free networks, and the Twitter network. It is found that compared with static networks, the opinion evolution in dynamic networks produces fewer opinion clusters and smaller opinion variances. The dynamic network mechanism reduces the weak relations between agents and improves the global clustering coefficient in the ER random networks but not in the Twitter network, which means that the network topology has an impact on results. Therefore, it is concluded that agents’ subjective behaviors significantly influence the outcome of opinion evolution and networks, which is consistent with real life.

Fuzzy preference relations (FPRs) have been widely used when ranking alternatives using pairwise comparison. However, when the FPRs have rank violations, the ranking results tend to contradict the decision makers’ (DMs) preferences. To date, no studies have examined the elimination of these FPR rank violations. Further, traditional consensus models utilizing distance measures between individual preferences are not suitable to manage classification-based consensus problems. To overcome these problems, this article develops several optimization models to address rank violations and h-rank consensus issues. First, the conditions that satisfy the FPRs’ preservation of order preferences (POPs) are analyzed and a system of constraints derived to ensure that the POPs are explicitly controlled by the optimization model, after which a mixed integer linear optimization model is developed to assist DMs to satisfy both the POP conditions and acceptable consistency. Finally, a linear model for the h-rank ($h \geq 2$) consensus reaching process is designed to ensure that each individual FPR also satisfies the POP conditions and acceptable consistency. The feasibility of the proposed models is illustrated using numerical examples, with extensive comparisons further validating the usefulness of the proposed models for group decision-making problems.

Recent studies have examined minimum cost consensus models from the moderator’s perspective. However, current dual models aimed at deriving a maximum return have been limited to special cases. A further problem has been the return allocation to the individuals. With this in mind, this article presents two general dual consensus models, the first of which is a minimum cost consensus model, and the second of which allows the different individuals to have different tolerance levels. In both models the individual opinions and the group opinion are connected using a weighted average operator. To understand the economic significance of these models, some properties, such as the weak duality property, the optimality property, and the strong duality property are given. The relationships between the individual opinions, the unit costs, and the unit returns are also proven, and the core return allocation problem is addressed. It was found that in both cases, the costs the moderator needed to pay to a given individual were precisely equal to the returns the individual deserved. Numerical case studies were used to illustrate the proposed models, with the simulation results providing insights into the practical use of these optimization consensus models.

In existing consensus models with optimization approaches, the modified preferences are often virtual values unrelated to the original rating scale. Furthermore, few optimization approaches have been developed to specifically deal with the consensus reaching process in uncertain multiple-criteria group decision making (MCGDM) problems. To investigate these issues, this paper is primarily interested in MCGDM optimization consensus models, in which the uncertain information can be represented using interval numbers. Integer optimization consensus models are established based on both the Manhattan distance and the Euclidean distance, and an antithetic-method-based particle swarm optimization algorithm is used to solve the nonlinear integer programming problems. Compared to existing approaches, the main characteristic of the proposed consensus models is that the generated preferences all belong to the original scales. A numerical example is given to validate the proposed models, and further simulation analyses shed light on the behavior of the developed models.

Individual consistency index is an important tool to solve decision making problems with multiplicative preference relations (also known as pairwise comparison matrices in the Analytical Hierarchy Process). This paper provides a new method to improve the consistency index, and the revised judgments for the decision maker belong to the original evaluation scale. An algorithm is designed to assist the decision maker in achieving a predefined consistency threshold. The developed approach does not depend on the consistency measure and the prioritization method. Some examples are given to illustrate effectiveness of the proposed approach. Comparing with the existing popular methods, the proposed one has two advantages: it has a smaller number of changes (in the two popular examples, only one element in the upper triangular preference relation needs to be revised); it is a heuristic approach which is much simpler than the nonlinear programming approach.

The main purpose of this paper is to propose a method to deal with multi-attribute group decision making (MAGDM) problem without any weight information of both decision makers and attributes under an interval type-2 fuzzy set (IT2 FS) environment where the decision information is provided with linguistic variables. First, we determine the weight of each attribute and decision maker based on entropy weight method and extended TOPSIS method, respectively, which are the objective weight determination methods and can decrease the effect of subjective consciousness. Then a MAGDM procedure based on the proposed weight determination methods and classical TOPSIS method under IT2 FSs environment is presented. In the framework, the linguistic decision information is represented by IT2 FSs. Finally, we apply the proposed MAGDM procedure to evaluate the harnessing of four rivers in Chengdu, the capital of Sichuan province, to illustrate the practicality and effectiveness of the proposed method.

In recent years, the public safety community has been aligning behind Long Term Evolution (LTE) technology as the basis for next generation public safety networks. Correspondingly, the 3rd Generation Partnership Project (3GPP) standards body has started multiple technical work initiatives on LTE enhancements to better support public safety use cases and requirements. Of particular importance to the public safety community is the addition of direct device-to-device (D2D) broadcast capability that can enable off-network push-to-talk (PTT) group communications in a manner equivalent to existing public safety communication systems. In this paper, we briefly explore the design objectives for, and challenges associated with, providing a D2D broadcast service. We then present and analyze a new distributed media access control (MAC) scheme for interference-aware coordination of both unicast and broadcast transmission. System-level simulation results are provided to validate the performance and the effectiveness of our approach, using a CSMA-based approach as a basis for comparison.

In group decision making (GDM) with qualitative settings, experts may require several possible linguistic values rather than a single term to express their preferences. A hesitant fuzzy linguistic term set has recently been developed to manage this situation. In line with this development, in this paper, we present a new framework model to address multiple attribute GDM with hesitant fuzzy linguistic information. First, the concept of a possibility distribution is defined. Based on the possibility distributions, some aggregation operators such as the hesitant fuzzy linguistic weighted average operator and the hesitant fuzzy linguistic ordered weighted average operator are proposed. A consensus measure is then defined and a consensus reaching process is given which uses different identification and direction rules compared with the existing methods. A selection process is also described to rank the alternatives. Both processes are necessary to support stakeholders when making rational decisions. Finally, two simulated examples are given to verify the practicability of the proposed approach.

The recently proposed hesitant fuzzy linguistic terms sets (HFLTSs) are utilized to represent the expert's subjective preferences in a linguistic preference relation and therefore a hesitant fuzzy linguistic preference relation (HFLPR) is constructed. This paper aims to present a consensus process to assist the experts in achieving a predefined consensus level in the case of HFLPRs. A possibility distribution based approach is introduced to deal with HFLTSs. Consensus degrees which assess the agreement among all the experts' preferences are defined on three levels: the pairs of alternatives level, the alternatives level and the preference relation level. A feedback mechanism based on the above consensus degrees is developed and the difference with the existing approach is discussed. The proposed consensus model is illustrated by a numerical example.

A fundamental aspect of group decision making is looking for approaches to reach consensus. This paper aims to propose a new approach to deal with the consensus reaching process for multiple attribute group decision making (MAGDM) with hesitant fuzzy linguistic information. The possibility distribution based approach is introduced to manage hesitant fuzzy linguistic term sets (HFLTSs). Based on the possibility distributions corresponding to the HFLTSs, a consensus degree for each expert is defined based on the distance between the individual decision matrix and the collective decision matrix. A simple procedure to reach a predefined consensus level is then developed where the feedback mechanism only uses the information of the consensus measure. A numerical example is provided to show how the computation of the proposed model goes.

In practical group decision making (GDM) problems adhere to uncertain and imprecise data, the decision makers may express their preferences using linguistic terms. The aim of this paper is to present a method to assist the consensus process and selection process of multi-criteria GDM (MCGDM) problem under linguistic setting. If the consensus level does not meet predefined requirements, an algorithm is provided to help the decision maker or moderator reach the consensus goal. Once the consensus reaching process is finished, the maximizing deviation method is used to derive the importance weights of the attributes. Then, the linguistic weighted arithmetic averaging (LWAA) operator of 2-tuple linguistic variables is used to obtain the overall assessment value of each alternative and the ranking order of all alternatives can be determined. Finally, one example of personal selection problem is given to show the use of the proposed method.

In this paper, we introduce a delay-sensitive service that involves transmitting large amounts of location-based data to nodes at multiple locations. Given a limited amount of access points (APs) and an abundance of service requests that result from the nodes moving around, a typical content delivery service would inevitably introduce considerable delay. To solve this problem, we analyze the movement pattern of mobile nodes and approximate it as a semi-Markov process. Based on this model, we explore different components of the underlying service delay and propose that APs should use a multicast strategy to minimize the queuing delay component. Furthermore, we demonstrate the feasibility of employing nodes, which already have their own local copies of location-relevant data, to relay such data to other nodes by employing one or multiple communication channels. Lastly, we examine the resulting algorithms and study their performance relative to baseline content-delivery schemes through simulations.

This paper presents an integrated routing and MAC scheduling protocol (IRMA) for multihop wireless mesh networks. The IRMA approach is motivated by the fact that the overall performance achieved by conventional layered approaches (802.11 MAC combined with independent ad hoc routing protocols) is significantly lower than the underlying network capacity. We propose to integrate the routing and MAC into a single protocol layer and use joint optimization techniques to establish end-to-end path and TDMA schedules for flows across the network. This approach achieves non-conflicting allocation of channel resources based on global or local traffic flow specifications and the network graph. The proposed method not only establishes interference-free MAC link schedules, but also helps to find optimal routes which can route around congested areas of the network. Two specific IRMA algorithms are proposed and evaluated in this paper. The first method solves min-hop routing, then optimizes link scheduling based on routing results and real-time flow demands. The second approach attempts to optimize routing and scheduling decisions simultaneously, using available MAC bandwidth information to route around congested areas. Both centralized and distributed algorithms based on these methods are proposed and evaluated with detailed simulations. Results show significant 2–3x improvements in network throughput when compared with baseline 802.11-based mesh networks using independent routing protocols.

Ad-hoc wireless networks with multi-rate radios (such as 802.11a, b, g) require a new class of MAC/PHY aware metrics that take into account factors such as physical-layer link speed and MAC-layer channel congestion. Conventional "layer 3" ad-hoc routing algorithms typically make routing decisions based on the minimum hop-count (MH). Use of the MH metric leads to the selection of paths with few hops, but one or more of these hops may turn out to be low-speed radio links due to adaptive rate selection at the physical layer We investigate a new cross-layer routing metric that takes into account both physical layer link speed and estimated channel congestion, thus aiming to minimize end-to-end delay that includes both transmission and access times. The proposed "PARMA" routing metric thus helps to spread the traffic across the "good links and nodes" in the network, increasing network capacity and reducing packet loss and delay. The paper presents the design and implementation of the proposed PARMA metric for proactive ad-hoc routing protocols, such as DSDV. DSDV modifications for incorporating the MAC/PHY aware metric into an ns-2 simulation model are given. Simulation results for typical multi-rate 802.11 ad-hoc network scenarios show that the proposed cross-layer PHY/MAC aware metric achieves significantly higher network throughput and decreases network congestion by selecting paths with high bit-rate links, while also avoiding areas of MAC congestion.

The paper presents a novel medium access control (MAC) protocol for QoS support in multi-hop ad-hoc wireless networks. The proposed D-LSMA (distributed link scheduling multiple access) protocol uses an extension of the 802.11 CSMA/CA procedure as the basis for a distributed link scheduling algorithm which results in dynamic TDMA-like bandwidth allocation among neighboring wireless nodes without the need for global synchronization. In addition to supporting QoS, the proposed scheduling technique also solves the "exposed node" problem in ad-hoc 802.11, thus resulting in improved throughput in many scenarios. Simulation results from an ns-2 model are presented for a 15-node random ad-hoc network. The results demonstrate significant performance improvements relative to ad-hoc 802.11, with capacity increases typically /spl sim/20% for the example considered. Also, the D-LSMA network is shown to offer far better real-time packet delay and fairness properties than 802.11, particularly under overload and heavy contention conditions.