With the growing number of publicly available Web APIs, developers often struggle to identify compatible and effective web API compositions to meet the requirements of complex application requirements. API composition recommendation has thus emerged as a key enabler for mashup development. However, most existing approaches rely solely on graph-based search algorithms over API correlation graphs, lacking a principled mechanism to assess the utility of candidate API compositions. To overcome this limitation, we propose OSWAR (Optimal Subset Web API Recommendation), a two-stage framework for API composition recommendation. In the recall phase, OSWAR identifies candidate compositions by performing a minimum Steiner tree search on an API correlation graph, ensuring functional completeness and compatibility. In the recommendation phase, we introduce an optimal subset oracle that learns a utility-based scoring function via variational inference to re-rank candidate compositions. Extensive experiments on the real-world ProgrammableWeb dataset demonstrate that OSWAR significantly outperforms several state-of-the-art baselines in terms of Precision, Recall, F1-score, Hit Rate, and mAP, particularly excelling when recommending Top-1 API composition.

Generative replay mitigates catastrophic forgetting by synthesizing samples from previously learned tasks and integrating them with data from the current task. While recent advances have improved generation fidelity and robustness, most existing approaches implicitly assume that replay is equally beneficial across samples, classes, or tasks. We revisit generative replay from a utility-guided perspective and ask two questions: which samples should be generated within each class, and how should the replay budget be allocated across past tasks? To this end, we study these questions in a feature-space replay setting and propose two complementary mechanisms. First, an intra-class utility module steers generation toward informative regions of the class distribution by considering uncertainty, representativeness, and diversity. Second, an inter-task allocation module adaptively distributes the replay budget across previously learned tasks based on measurable signals of forgetting and interference. Experiments on CIFAR-100 and ImageNet-100 under different class-incremental protocols show that the proposed method consistently improves incremental accuracy and reduces forgetting. Further analysis suggests that utility-guided replay better preserves class structure and maintains more stable decision boundaries.

We design and demonstrate the first field-operational GraphRAG agent for real-time information query in large-scale hierarchical optical networks, achieving 96.25% Natural Language-to-Gremlin accuracy via Knowledge Graph General Rules and iterative optimization.

With the rapid development of ubiquitous networks and unmanned devices, air quality monitoring data are increasingly collected via wireless networks from sensors at multiple monitoring stations. However, the complexity of these data introduces significant challenges. Existing spatiotemporal methods for air quality prediction often struggle with issues such as inadequate handling of spatial relationships, difficulties in modeling long-term temporal dependencies, and limited generalization capabilities. To address these challenges, this article proposes a novel spatiotemporal modeling approach-spatiotemporal lambda -Bernstein graph convolutional network with transformer for multisite air quality prediction (ST-BernT). This method constructs a graph structure based on spatiotemporal correlations, integrating a Gaussian-weighted adjacency matrix derived from geographic distances with an adjacency matrix capturing the temporal correlations of pollutant concentration time series, thereby precisely modeling spatial dependencies. Subsequently, a dynamic filter adjusted by lambda -Bernstein polynomials is proposed to adaptively process spatiotemporal data characterized by the coexistence of low- and high-frequency components on the graph. Furthermore, a hierarchical generative transformer (GPHT) is introduced to enhance the model's ability to capture long-term temporal patterns, such as periodicity and seasonality, while supporting parallel prediction across multiple sites, significantly improving computational efficiency and accuracy. Experimental results demonstrate that ST-BernT exhibits strong accuracy, adaptability, and generalization capability in multisite air quality prediction tasks, particularly showing enhanced robustness in large-scale long-term forecasting scenarios.

Developers integrate web Application Programming Interfaces (APIs) into edge applications, enabling data expansion to the edge computing area for comprehensive coverage of devices in that region. To develop edge applications, developers search API categories to select APIs that meet specific functionalities. Therefore, the accurate classification of APIs becomes critically important. However, existing approaches, as evident on platforms like programableweb.com, face significant challenges. Firstly, sparsity in API data reduces classification accuracy in works focusing on single-dimensional API information. Secondly, the multidimensional and heterogeneous structure of web APIs adds complexity to data mining tasks, requiring sophisticated techniques for effective integration and analysis of diverse data aspects. Lastly, the long-tailed distribution of API data introduces biases, compromising the fairness of classification efforts. Addressing these challenges, we propose MDGCN-Lt, an API classification approach offering flexibility in using multi-dimensional heterogeneous data. It tackles data sparsity through deep graph convolutional networks, exploring high-order feature interactions among API nodes. MDGCN-Lt employs a loss function with logit adjustment, enhancing efficiency in handling long-tail data scenarios. Empirical results affirm our approach's superiority over existing methods.

Air quality predictions play a critical role in shaping governmental policies and measures worldwide, aimed at mitigating the health risks posed by air pollution to the public. However, current spatiotemporal prediction methods frequently encounter challenges in accurately extracting spatial features, which adversely affects the overall predictive performance. To mitigate this limitation, we introduce AGTCN, a novel air quality forecasting model that integrates an adaptive gating mechanism. The model leverages an enhanced Graph Convolutional Network (GCN), wherein the adaptive gates dynamically regulate spatial feature extraction based on geographic inputs. Subsequently, the refined spatial representations are fused with a Temporal Convolutional Network (TCN) to model the joint spatiotemporal dynamics. Ultimately, the synthesized features are passed through a prediction layer to generate the final outputs. We validated the model's performance using a real dataset, and both comparative experiments and ablation studies demonstrated the model's effectiveness.

In recent years, with the development of cyber-physical-social systems (CPSSs), the integration of the Internet of Things (IoT) and social networks has been advocated to promote further development. Multimedia big data (MMBD) generated by Social Internet of Things (SIoT) devices is commonly distributed and stored on edge servers. In this context, collaborative processing of user requests among edge servers becomes crucial to meet the low-latency retrieval needs of users. However, in practice, the lack of global data storage information leads to significant time and space costs when identifying servers across the entire system. To tackle this issue, we propose a distributed edge data indexing system named FMQS, aiming to achieve fast MMBD queries at the edge. First, we introduce the cuckoo filter (CF) tree index structure, which enables rapid edge data queries. Second, we utilize a novel indexing structure based on the grouping cuckoo bloom filter (GCF) tree to improve query performance. What is more, we conceptualize distributed edge servers as entities within social networks, optimizing their relationships in terms of geographical proximity to improve query performance. Finally, we conduct performance evaluation, security verification, and analysis of the FMQS scheme on an edge storage system consisting of 90 edge servers, further confirming the effectiveness and feasibility of FMQS in MMBD queries.

We propose a novel high performance computing (HPC) network architecture $\mathrm{HFOS}_L$ based on $L$ parallel levels distributed low radix fast optical switches (FOS). We provide a detailed description of the blade, FOS and the operation of the $\mathrm{HFOS}_L$ network. The $\mathrm{HFOS}_L$ HPC network is highly scalable, and $\mathrm{HFOS}_4$ architecture can support an extremely large HPC network of 65,536 blades under distributed FOS with the same radix of 16 in each level. In principle, the $\mathrm{HFOS}_L$ HPC network can be built by FOSes with different radices at each level. To find out the best configuration of FOS at each level and solve the energy and cost optimization problem in $\mathrm{HFOS}_L$ network, we break down all the components in the FOS and develop the energy and cost models for the FOS. We verify that the energy and cost per radix functions of FOS are convex functions. Given this foundation, the theoretical investigation of the energy and cost optimization problem shows that the $\mathrm{HFOS}_L$ network could achieve the minimum energy and cost only when the FOS radices of all levels in $\mathrm{HFOS}_L$ network are the same. Besides, the cost and power consumption of $\mathrm{HFOS}_L$ networks are compared with a widely used Leaf-Spine network.

Optical networks have been evolving from proprietary and close systems to open line systems, and further to open box systems. Technologies that enabled the evolution are reviewed and discussed.

With the popularity of 5G and the rapid development of mobile terminals, an endless stream of short video software exists.Browsing short-form mobile video in fragmented time has become the mainstream of user's life.Hence, designing an efficient short video recommendation method has become important for major network platforms to attract users and satisfy their requirements.Nevertheless, the explosive growth of data leads to the low efficiency of the algorithm, which fails to distill users' points of interest on one hand effectively.On the other hand, integrating user preferences and the content of items urgently intensify the requirements for platform recommendation.In this paper, we propose a collaborative filtering algorithm, integrating time context information and user context, which pours attention into expanding and discovering user interest.In the first place, we introduce the temporal context information into the typical collaborative filtering algorithm, and leverage the popularity penalty function to weight the similarity between recommended short videos and the historical short videos.There remains one more point.We also introduce the user situation into the traditional collaborative filtering recommendation algorithm, considering the context information of users in the generation recommendation stage, and weight the recommended short-form videos of candidates.At last, a diverse approach is used to generate a Top-K recommendation list for users.And through a case study, we illustrate the accuracy and diversity of the proposed method.

Software developers create mashups of their interest by searching, selecting, and combining web Application Programming Interfaces (APIs) with diverse functionalities. However, achieving the selection of APIs relies on a crucial prerequisite: accurate classification of APIs. However, the traditional web API classification methods are often based on a single piece of API function's information such as manual annotations and description documents which decreases the accuracy of API classification results significantly, especially for the web API sharing communities with sparse data (e.g., programmeableweb.com). To tackle this issue, we propose BDGCN which improves some existing web API classification approaches by adding isomorphic graph of strong correlation among different web APIs on the basis of API description documents. Concretely, at the first, we integrate the information of multiple dimensions of web APIs as embedded inputs in which we then build an API-API strongly correlated isomorphic network to strengthen the association among API individuals, and to initialize the embedding vector from the description of each API by tuning the pre-trained Bidi-rectional Encoder Representation from Transformers (BERT) model. Besides, in order to smooth the impact of high-order features between neighboring nodes, a deep graph convolutional network is introduced in this paper that is consistent with the problem. In fact, we set up a categorization layer and utilize cross-entropy function to achieve accurate classification of APIs. Experimental results demonstrate the superiority of our method compared to the existing methods.

Open and disaggregated systems supporting flexgrid operation and ROADM are being introduced to data center interconnect networks. System integration and network automation are effective to further increase the network efficiency in the future. © 2022 The Author(s)

A spectrum optimization method without service interruption is demonstrated in a flexgrid disaggregated system by extending a transponder laser bright tuning range, adapting OpenConfig data models in devices and implementing algorithms in a network management system. The whole system is composed by commercialized devices.

As the fundamental infrastructure of the Internet, the optical network carries a great amount of Internet traffic. There would be great financial losses if some faults happen. Therefore, fault location is very important for the operation and maintenance in optical networks. Due to complex relationships among each network element in topology level, each board in network element level, and each component in board level, the con?crete fault location is hard for traditional method. In recent years, machine learning, es?pecially deep learning, has been applied to many complex problems, because machine learning can find potential non-linear mapping from some inputs to the output. In this paper, we introduce supervised machine learning to propose a complete process for fault location. Firstly, we use data preprocessing, data annotation, and data augmenta?tion in order to process original collected data to build a high-quality dataset. Then, two machine learning algorithms (convolutional neural networks and deep neural networks) are applied on the dataset. The evaluation on commercial optical networks shows that this process helps improve the quality of dataset, and two algorithms perform well on fault location.

We design and implement a disaggregated ROADM network by treating an optical multiplex section as a basic building block. Both automatic channel provisioning and end-to-end link power adjustment are demonstrated in such network.

The emergence of edge computing provides an effective solution to execute distributed model training (DMT). The deployment of training data among edge nodes affects the training efficiency and network resource usage. This letter aims for the efficient provisioning of DMT services by optimizing the partition and distribution of training data in edge computing-enabled optical networks. An integer linear programming (ILP) model and a data parallelism deployment algorithm (DPDA) are proposed to solve this problem. The performance of the proposed approaches is evaluated through simulation. Simulation results show that the proposed algorithm can deploy more DMT services compared with benchmark.

With the increasing bandwidth requirements for people and the development of urbanization, the movement of the population in the city (especially the supercity) has an increasing influence on the traffic distribution in both space and time dimensions. The imbalanced distribution results in the regional blocking of different areas in different time periods and reduces the spectrum resource utilization in elastic optical networks. To resolve this problem, this paper proposes a tidal traffic model to formulate a kind of tidal traffic phenomenon firstly. Based on the analysis for this model, the area-aware routing and spectrum allocation algorithm that focuses on the traffic adjustment in specific functional areas is proposed. And two benchmark algorithms named min-hop k-shortest path routing algorithm and occupied-slots-as-weight k-shortest path routing algorithm are introduced. The evaluation results show that, compared to the benchmark algorithms, the proposed area-aware algorithm could reduce the blocking probability efficiently from 2% to 47% with low time complexity.

Traffic flows are often processed by a chain of Service Functions (SFs) (known as Service Function Chaining (SFC)) to satisfy service requirements. The deployed path for a SFC is called Service Function Path (SFP). SFs can be virtualized and migrated to datacenters, thanks to the evolution of Software Defined Network (SDN) and Network Function Virtualization (NFV). In such a scenario, provisioning of paths (i.e., SFPs) between virtualized network functions is an important problem. SFP provisioning becomes more complex in a multi-domain network topology. ‘Topology aggregation’ helps to create a single-domain view of such a network by abstracting multi-domain networks. However, traditional ‘topology aggregation’ methods are unable to abstract SF resources properly, which is required for SFP provisioning. In this paper, we propose an SFC-Oriented Topology Aggregation (SOTA) method to enable abstraction for SFs in multi-domain optical networks. This study explores the node and the link aggregation degree to evaluate information compression during the ‘Topology aggregation’ process. Additionally, we also propose a new data structure named wheel matrix and related operations to store routing information in the aggregated topology. Based on SOTA, we propose two cross-domain SFP provisioning algorithms named Ordered Anchor Selection (OAS) and ${k}$ -paths OAS (K-OAS), and a benchmark named Global OAS (GOAS). Simulation results show that SOTA could aggregate large-scale multi-domain optical networks into a small network that contains only 6.9% of the nodes and 10.1% of the links. Both OAS and K-OAS can calculate SFPs efficiently and reduce blocking probability up to 52.10% compared to the benchmark.

This article reports a technical demonstration showing the use of the ONOS SDN controller for disaggregated transport networks, summarizing the latest developments and results within the Open Networking Foundation (ONF) Open Disaggregated Transport Networks (ODTN) project. The demonstration mainly covers the dynamic provisioning of data connectivity services and advanced automatic failure recovery, both at the control and data plane levels. For the provisioning, we demonstrate the usage of open and standard protocols and interfaces. For the recovery part, we first demonstrate, covering the control plane, how ONOS can behave as a logically centralized controller while using multiple coordinated instances for robustness. We show how the different devices remain under control even in the event of a failure of one of such instances, relying on the ATOMIX framework and the dynamic real-time negotiation of device mastership. For the data plane, we demonstrate the capabilities of the controller to perform automatic restoration of optical services.