Graph Neural Networks (GNNs) offer superior modeling capabilities for text classification by capturing complex spatial features within semantic representations. However, existing graph-based approaches often suffer from computational inefficiency and limited ability to model both fine-grained local structures and the sequential nature of text. To address these challenges, we propose HC2-GNN, a Hierarchical Clustering and Coarsening Graph Neural Network, which introduces a novel lightweight graph clustering algorithm called Compromise Conductance Graph Clustering (C2GC). C2GC enables efficient graph clustering while simultaneously preserving both the textual order and the topological coherence of subgraphs. Furthermore, it incorporates a virtue cluster mechanism that expands each subgraph with semantically relevant neighbors, explicitly enabling cross-cluster information propagation without compromising local structural integrity. HC2-GNN aggregates local and global features by combining subgraph-level and full-graph representations, enhancing semantic discriminability for classification. Extensive experiments on benchmark datasets demonstrate that HC2-GNN consistently outperforms existing state-of-the-art text classification methods.

Affective Explanation Captioning (AEC) aims to perform viewer-centered visual emotion analysis by not only identifying the emotions evoked by an image but also explaining their underlying causes. Prior efforts have achieved promising results by fine-tuning LLMs on affective data; however, two key challenges remain: 1) the inherent subjectivity of human emotion leads to diverse interpretations of the same image, making it difficult for models to catch dominant emotions; and 2) the affective gap between abstract emotions and concrete visual content hinders models from capturing both semantic and emotional aspects effectively. To tackle these challenges, we propose Consensus-Prompted Emotion Reasoning (CPER), a new framework that explicitly models emotional diversity and enforces emotional-semantic alignment. Inspired by psychological studies, we observe that common emotional patterns often emerge within certain groups, which we refer to as affective consensus. Capturing this consensus across varying levels is helpful for bridging the subjectivity in AEC. Specifically, we introduce a consensus-based bucket prompt, which depicts the consensus level of each emotional perspective, serving as a control signal to adjust the emotion reasoning. To reconcile abstract emotion understanding and concrete visual grounding, we design a dual-space representation, where a CLIP encoder extracts objective semantic evidence and an emotion encoder captures abstract affective cues for AEC. Furthermore, an emotion consistency learning strategy is devised, which explicitly aligns the generated explanation with the input image and the emotion label, ensuring both emotionally and semantically grounded explanations. Extensive experiments on three benchmark datasets, ranging from visual arts (ArtEmis v1.0 and ArtEmis v2.0) and real-world images (Affection), demonstrate the effectiveness of our CPER in terms of emotional diversity and semantic coherence compared to state-of-the-art methods. Our code is publicly available at https://github.com/songpipi/CPER.

Efficient production scheduling of aviation components is essential for optimizing manufacturing equipment and resource utilization. Traditional scheduling requires experts to develop complex models and optimization algorithms, making the process time-consuming and expertise-dependent. This paper proposes LLM-Scheduling, a multi-agent framework where each agent, powered by a large language model (LLM), autonomously executes scheduling tasks based on natural language inputs. The framework integrates retrieval-augmented generation (RAG) to provide relevant domain knowledge and employs a Hybrid-Search strategy to enhance optimization performance. Experimental results demonstrate that LLM-Scheduling outperforms conventional approaches in most cases, achieving more efficient scheduling and improved adaptability in dynamic manufacturing environments. Moreover, the framework shows promise in enhancing energy efficiency by optimizing machine utilization and reducing idle time. These findings suggest that LLM-driven scheduling can significantly contribute to smarter and more sustainable manufacturing processes.

The rapid advancement of Software-Defined Networks (SDN) and Network Function Virtualization (NFV) has popularized the adoption of the Service Function Chain (SFC) paradigm for efficient network service delivery. This paradigm leverages the flexibility and cost-effectiveness of deploying Virtual Network Functions (VNFs) as software entities or virtual machines on off-the-shelf servers. Chaining VNFs together allows traffic to be directed through the network as required. However, existing algorithms for traffic steering and routing path computation in SFC suffer from many challenges, including complexity, lack of scalability, and low time efficiency. This paper focuses on addressing the challenges associated with VNF placement and SFC chaining in SDN/NFV-enabled networks. Our objective is to identify an optimal solution for VNF placement that maximizes the utilization of network resources. We formulate the problem as a Binary Integer Programming (BIP) model to accomplish this. Additionally, we propose a novel algorithm called DeepSelector, which incorporates deep learning techniques and an intelligent node selection network to determine the optimal placement of VNFs for SFC requests. Through performance evaluation, we demonstrate that DeepSelector achieves high network resource utilization and offers efficient VNF placement computation, significantly improving overall network performance.

The emergence of Software-Defined Networks (SDN) and Network Function Virtualization (NFV) has made Service Function Chain (SFC) a popular method for delivering network services. This innovative computing and networking paradigm allows Virtual Network Functions (VNFs) to be cost-effectively deployed on a network of physical equipment flexibly and elastically. Traffic can be directed as needed by linking VNFs as an SFC. However, the current algorithms for VNF placement computation and traffic steering in SFC are often complex, unscalable, and time-consuming. This paper investigates the VNF placement and SFC chaining problem in NFV-enabled networks. To obtain the VNF placement solution that maximizes network resource utilization, we formulate the problem as a Binary Integer Programming (BIP) model. Additionally, we introduce a novel Deep Learning-based VNF Placement Algorithm (DLVPA) that uses an intelligent node selection network to place VNFs for SFC requests. Performance evaluations demonstrate that DLVPA can effectively improve network resource utilization and achieve high solution computation time efficiency.

With the emergence of Network Function Virtualization (NFV) and Software-Defined Networks (SDN), Service Function Chaining (SFC) has evolved into a popular paradigm for carrying and fulfilling network services. This new networking and computing paradigm enables virtual network functions (VNFs) to be placed in virtual machines/software entities on a network of physical devices elastically and flexibly with lower capital and operating expenditures. However, for cloud service providers, how to migrate VNFs in NFV-enabled networks for more flexible services is a critical issue that needs to be addressed. Currently, research on VNF migration mainly focuses on how to migrate a single VNF while ignoring the VNF sharing and concurrent migration. This paper assumes that each placed VNF can serve multiple SFCs. We focus on selecting the best migration location for concurrently migrating VNF instances based on actual network conditions. First, we formulate the VNF migration problem as an optimization model whose goal is to minimize the end-to-end delay of all influenced SFCs while guaranteeing network load balance after migration. Next, we design a Two-Stage Hybrid Genetic Evolution (T-SHGE) solution to solve the VNF migration problem. Finally, we combine previous experimental data to generate realistic VNF traffic patterns and evaluate the algorithm. Simulation results show that the SFC delay after migration calculated by T-SHGE is close to the optimal results and much lower than the benchmarks. In addition, it effectively guarantees the load balancing of the network after migration.

Source free domain adaptation (SFDA) transfers a single-source model to the unlabeled target domain without accessing the source data. With the intelligence development of various fields, a zoo of source models is more commonly available, arising in a new setting called multi-source-free domain adaptation (MSFDA). We find that the critical inborn challenge of MSFDA is how to estimate the importance (contribution) of each source model. In this paper, we shed new Bayesian light on the fact that the posterior probability of source importance connects to discriminability and transferability. We propose Discriminability And Transferability Estimation (DATE), a universal solution for source importance estimation. Specifically, a proxy discriminability perception module equips with habitat uncertainty and density to evaluate each sample's surrounding environment. A source-similarity transferability perception module quantifies the data distribution similarity and encourages the transferability to be reasonably distributed with a domain diversity loss. Extensive experiments show that DATE can precisely and objectively estimate the source importance and outperform prior arts by non-trivial margins. Moreover, experiments demonstrate that DATE can take the most popular SFDA networks as backbones and make them become advanced MSFDA solutions.

Source free domain adaptation (SFDA) aims to transfer a trained source model to the unlabeled target domain without accessing the source data. However, the SFDA setting faces a performance bottleneck due to the absence of source data and target supervised information, as evidenced by the limited performance gains of the newest SFDA methods. Active source free domain adaptation (ASFDA) can break through the problem by exploring and exploiting a small set of informative samples via active learning. In this paper, we first find that those satisfying the proper-ties of neighbor-chaotic, individual-different, and source-dissimilar are the best points to select. We define them as the minimum happy (MH) points challenging to explore with existing methods. We propose minimum happy points learning (MHPL) to explore and exploit MH points actively. We design three unique strategies: neighbor environment uncertainty, neighbor diversity relaxation, and one-shot querying, to explore the MH points. Further, to fully exploit MH points in the learning process, we design a neighbor focal loss that assigns the weighted neighbor purity to the cross entropy loss of MH points to make the model focus more on them. Extensive experiments verify that MHPL remarkably exceeds the various types of baselines and achieves significant performance gains at a small cost of labeling.

Network service providers benefit greatly from Network Function Virtualization (NFV), which allows them to outsource their Network Functions (NFs) to cloud data centers flexibly. This paper focuses on the Virtual Network Function (VNF) placement in cloud data centers while maximizing the network’s accepted Service Function Chain Requests (SFCRs). To optimize resource utilization, we consider two key factors that are often overlooked: time-varying workloads and VNF sharing based on multi-tenancy technology. We formulate the VNF placement problem as an Integer Linear Programming (ILP) model. To solve the ILP, we devise a Throughput Optimization Heuristic Solution (TOHS). Finally, we conduct a detailed numerical simulation and compare our results with contrasting schemes in the existing literature. Our evaluation shows that the performance of TOHS is near to results derived by ILP solver for small-scale problems. In addition, TOHS outperforms other solutions in various scenarios, resulting in higher network throughput and better utilization of network resources.

The Service Function Chain (SFC) has become a popular paradigm to complete mobile services due to the advancements in Mobile Edge Computing (MEC) and Network Function Virtualization (NFV). This new computing and networking paradigm allows Virtual Network Functions (VNFs) to be placed in physical devices within MEC-NFV networks cost-effectively and flexibly. However, most existing VNF placement algorithms are complex, unscalable, and time-consuming. In this paper, we investigate the VNF placement problem in MEC-NFV networks and formulate an optimization model to optimize network resource utilization. We introduce a novel Deep Learning-based VNF Placement Approach (DLVPA) that intelligently selects nodes and places VNFs for SFC requests. Performance evaluations demonstrate that DLVPA can effectively improve network resource utilization.

In the 6G vision, networks are expected to be more flexible in quickly solving network traffic scheduling issues and deploying services. Network Function Virtualization (NFV) is an innovative technology that involves extracting network functions from dedicated equipment to create Virtual Network Functions (VNFs). These VNFs are then chained together to form a Service Function Chain (SFC) that provides network service. However, there are still some issues with existing network service orchestration tools, such as unreasonable multi-traffic scheduling and additional programming requirements for end-users. We have developed a solution to address the challenges posed by data coupling and bandwidth preemption in multi-service environments. Our dynamic Quality of Service (QoS) Guarantee model utilizes hierarchical analysis to prioritize traffic among multiple service data streams and employs a service scheduling algorithm based on a weighted fair queue to allocate link resources. For user convenience, we have also created an intuitive web orchestration platform called EasyOrchestrator, enabling users to encapsulate common VNFs and build services quickly. Our experimental evaluation has shown that EasyOrchestrator significantly reduces service construction time compared to the benchmark. At the same time, our QoS assurance mechanism effectively minimizes network congestion and ensures the successful operation of high-priority services.

Virtual Network Function (VNF) placement in NTNs is challenging because Non-Terrestrial Networks (NTNs), such as satellite networks, have limited resources regarding computational power and rate. However, existing solutions do not consider satellites’ resource constraints and the bandwidth constraints of links, which are essential metrics for designing VNF placement strategies in NTNs. Utilizing Network Function Virtualization (NFV) technology to deploy related network services on satellites in VNFs is a reasonable way. This paper focuses on delay-aware VNF placement in 6G NTNs to meet the ultra-low delay requirements of different applications. In addition, we also consider how to improve the resource utilization of servers to eliminate the resource bottlenecks of resource-constrained 6G NTN facilities. Then we formulate the VNF placement problem as a weighted graph-matching problem, aiming to maximize resource utilization. We propose the Linear Programming based algorithm and the Hungarian-based algorithm to solve the VNF placement problem. Evaluation results show that our proposed solutions outperform the benchmarks regarding resource utilization and execution time.