Through exploiting decentralized data from multi-source Internet-of-Things (IoT) devices, federated learning (FL) can accomplish the training of deep neural network (DNN) models in a privacy-preserving manner to provide premium intelligent services. Due to portability considerations, most IoT devices are computing-constrained which cannot afford frequent DNN model training in FL. Existing approaches use model compression techniques to reduce computing cost of IoT devices, whereas accuracy degradation is inevitably incurred. To address this challenge, we propose an elastic federated learning collaboration framework, namely EFLCF, to accommodate limited computing resources of IoT devices. Specifically, we first design an FL-oriented elastic neural network model with multiple-width subnets, and couple it with an FL device-server collaboration framework to form EFLCF, thereby releasing computing cost pressure of IoT devices. We then develop a freezing-assisted wide-to-narrow training mechanism to realize efficient device-server distributed training and further reduce device computing cost. Finally, we design an entropy-based narrow-to-wide elastic inference mechanism to decrease computing cost of inference without compromising accuracy. Experiments demonstrate that compared to well-known benchmarks, our EFLCF can reduce up to 97.65% device computing cost and improve up to 48.3% accuracy in training, while reducing up to 42.5% computing cost in inference.

The accumulation of large-scale multi-omics datasets has brought new opportunities for precise disease treatment. However, the inherent complexity of inter- and intra-omics relationships presents considerable obstacles to the precise integration of multi-omics data. Here, we propose a Hierarchical Feature Aggregation and Cross-Modal Attention Fusion (HCMAF) framework to integrate multi-omics data for patient classification and biomarker identification. Specifically, to capture both the specific information inherent in each omics data and complex cross-omics interactions, HCMAF incorporates three innovative modules. The hierarchical feature aggregation graph attention (HGAT) module captures intra-omics topological features through adaptive neighborhood aggregation. The cross-modal attention (CMA) module pinpoints inter-omics complementarity by modeling cross-omics dependencies. Finally, the confidence-driven multi-omics fusion (CMF) module dynamically integrates omics-specific predictions through learnable reliability weights. Comprehensive experiments on four public benchmark datasets show that HCMAF achieves better classification performance and consistently surpasses leading existing methods. Further component analysis confirms the crucial role of the HGAT, CMA and CMF modules in ensuring overall model effectiveness.

Web APIs have become a fundamental paradigm in the Web 4.0 era, with mashup services emerging as a transformative technology that combines multiple APIs to create comprehensive services. However, existing approaches exhibit two significant limitations: overlooking the quality and completeness of recommendation contexts of new mashup requirements, and failing to effectively extract high-quality collaborative features from multi-perspective service relationships. To address these limitations, we propose LMSR, a novelLLM-enhancedMulti-perspectiveService Feature Learning framework for Web APIRecommendation. LMSR first leverages general-purpose LLM to refine and encode the original requirement descriptions, and employs a Mixture of Service Experts (MoSE)-based context prediction model to precisely predict service information relevant to new requirements, establishing a comprehensive and high-quality recommendation context for mashup requirements. Furthermore, by integrating the predicted recommendation contexts into the LLM through fine-tuning, LMSR effectively extracts collaborative features from multi-perspective service relationships between mashup requirements and APIs, ultimately achieving precise Web API recommendation. Comprehensive experiments on real-world datasets demonstrate that LMSR significantly outperforms 11 baseline approaches across precision, recall, F1-score, and NDCG, validating its effectiveness in Web API recommendation. The codes are available athttps://scdm-shu.github.io/codes/LMSR.zip.

While Large Language Models (LLMs) show promise for Text-Attributed Graphs (TAGs) learning, their deployment is hindered by computational demands. Graph Neural Networks (GNNs) are efficient but struggle with TAGs' complex semantics. We propose LinguGKD, a novel LLM-to-GNN knowledge distillation framework that enables transferring both local semantic details and global structural information from LLMs to GNNs. First, it introduces TAG-oriented instruction tuning, enhancing LLMs with graph-specific knowledge through carefully designed prompts. Next, it develops a layer-adaptive multi-scale contrastive distillation strategy aligning LLM and GNN features at multiple granularities, from node-level to graph-level. Finally, the distilled GNNs combine the semantic richness of LLMs with the computational efficiency of traditional GNNs. Experiments demonstrate that LinguGKD outperforms existing graph distillation frameworks, the distilled simple GNNs achieve comparable or superior performance to more complex GNNs and teacher LLMs, while maintaining computational efficiency. This work bridges the gap between LLMs and GNNs, facilitating advanced graph learning in resource-constrained environments and providing a framework to leverage ongoing LLM advancements for GNN improvement.

Edge intelligence, the fusion of edge computing and artificial intelligence (AI), drives the advancement of intelligent Internet of Things (IoT). Since AI applications are often data-and computation-intensive, resource-scarce edge devices need to migrate data to resource-rich edge servers through computation offloading to meet requirements such as energy efficiency and low latency. Existing studies often focus on CPU-based edge systems and neglect the impacts of other components, such as memory, on offloading. From a parallel processing perspective, this article establishes a system model and a multi-objective optimization model for edge intelligence systems with heterogeneous components, including diverse processors, memory, network, and applications, to minimize system energy consumption, total execution time, and the workload ratio of edge servers. A multi-objective optimization algorithm integrating archive initialization, hybrid perturbation, clustering, and modified simulated annealing is proposed and validated through experiments using real-world software and hardware. Results demonstrate that the proposed algorithm significantly outperforms comparative algorithms in terms of inverted generational distance, pure diversity, and run-time while revealing the influence of application characteristics on offloading performance.

Edge service deployment has attracted significant attention in recent years, aiming to optimize service placement on edge servers while satisfying diverse requirements. However, existing approaches often overlook the influence of geographical contexts on service demands, where user needs vary significantly across regions with distinct characteristics. They also fail to account for differences between direct responses and multi-hop forwarding in edge network topology, leading to unsatisfactory edge service deployment strategies. To this end, we formulate the Points of Interest-Based Edge Service Deployment (POI-ESD) problem with topology-aware optimization, integrating POI attributes and spatial distributions while incorporating edge network topology to enhance service placement. By proving the$\mathcal{N}P$-hardness of POI-ESD problem, we propose a novel graph-encoded genetic algorithm, MTGA, to efficiently generate high-quality deployment strategies. It ensures strategic placement of edge services in regions that best match user demands, improving the service utilization and satisfiability for edge users. Extensive experiments on a real-world dataset combining Shanghai Telecom and Baidu Maps POI data demonstrate that MTGA significantly outperforms existing competing approaches, achieving superior performance of edge service deployment.

Federated learning (FL), an emerging data-secure distributed training paradigm, unites massive isolated Internet of Things (IoT) device nodes to collaboratively train a global neural network (NN) model without the exposure of their local multimedia data. However, constrained by the synchronous NN model integration nature of FL, there is a training latency inconsistency among heterogeneous devices, which significantly deteriorates FL training efficiency. Meanwhile, frequent local NN training and transmission impose high energy consumption pressure on users. To tackle these issues, this paper proposes a premium multi-width NN-assisted hierarchical FL (HFL) framework in heterogeneous cloud-edge-device computing to achieve remarkable training speedup and energy conservation. Specifically, a heterogeneity-aware NN width coefficient determination algorithm, which flexibly assigns a subnet with a suitable width to each user device based on its computing ability, is first applied to shorten the HFL training latency. Subsequently, to integrate subnets with different width topologies, we design a width-aware adaptive NN model integration approach to effectively ensure the accuracy of the integrated global NN model. Finally, a latency-aware energy saving strategy is introduced to reduce energy consumption. Experimental results demonstrate that our proposed framework outperforms state-of-the-art benchmarks, and attains up to 42.42% enhancement in accuracy, 81.5% reduction in training latency, and 40.9% optimization in energy cost.

Connected vehicles due to the high mobility and dynamic network topologies of connected vehicles require accurate QoS that includes high throughput and low latency to assess satisfactory QoE. Existing methods mainly focus on centralized QoS prediction while paying little attention to distributed mobile QoS prediction, making it challenging to protect user privacy information when invoking Web services. Moreover, even though some advanced centralized methods can be transformed into federated architectures, they often face difficulty in capturing latent feature representations of users and services and learning personalized prediction layers between them due to the heterogeneity of the QoS dataset. To address the above issues, we propose a novel framework for distributed QoS prediction, called Combining Personalized Federated Hypernetworks and Shared Residual Learning for Distributed QoS Prediction (FHR-DQP) . FHR-DQP adopts the federated averaging (FedAvg) to aggregate location-aware residual shared feature information across all clients. Additionally, a hypernetwork is leveraged to generate personalized networks for user-service QoS prediction in each client. These components are integrated as a hybrid framework that performs training using a federated approach and makes personalized QoS predictions within each client. Extensive experiments are conducted on a real-world benchmark QoS dataset called WS-DREAM, containing nearly 2,000,000 historical QoS invocation records. Compared with both centralized and federated competing baselines, the results demonstrate that FHR-DQP achieves the highest performance for distributed QoS prediction, when it provides privacy-preserving of users’ QoS invocations.

Accurate prediction of temporal QoS is crucial for maintaining service reliability and enhancing user satisfaction in dynamic service-oriented environments. However, current methods often neglect high-order latent collaborative relationships and fail to dynamically adjust feature learning for specific user-service invocations, which are critical for precise feature extraction within each time slice. Moreover, the prevalent use of RNNs for modeling temporal feature evolution patterns is constrained by their inherent difficulty in managing long-range dependencies, thereby limiting the detection of long-term QoS trends across multiple time slices. These shortcomings dramatically degrade the performance of temporal QoS prediction. To address the two issues, we propose a novel Graph Attention Collaborative Learning (GACL) framework for temporal QoS prediction. Building on a dynamic user-service invocation graph to comprehensively model historical interactions, it designs a target-prompt graph attention network to extract deep latent features of users and services at each time slice, considering implicit target-neighboring collaborative relationships and historical QoS values. Additionally, a multi-layer Transformer encoder is introduced to uncover temporal feature evolution patterns, enhancing temporal QoS prediction. Extensive experiments on the WS-DREAM dataset demonstrate that GACL significantly outperforms state-of-the-art methods for temporal QoS prediction across multiple evaluation metrics, achieving the improvements of up to 38.80%.

How to efficiently deploying the service components of a data-intensive application on cloud and edge servers to minimize its latency is one of the main challenges for service providers. Most existing studies consider either service deployment or data placement, rather than their joint optimization. This work considers the driving relationship between data and services in a heterogeneous environment including remote cloud and nearby edge servers, and aims to obtain a desired data placement and service deployment scheme while meeting user requirements for service quality. Firstly, we formulate the problem and decouple data placement from service deployment by polynomial reduction. Then, a priority-based data placement strategy is proposed, which can generate a data placement scheme. After that, the original problem is transformed into a classical assignment problem, and a service deployment strategy based on an improved Hungarian algorithm is proposed to obtain a service deployment scheme. Then, a dynamic adjustment strategy based on response weight is proposed to dynamically adjust the data placement and service deployment scheme in order to reduce response latency, and obtain the final scheme. Finally, a series of comparative experiments were conducted, pitting our algorithms against several baseline and SOTA algorithms. The results show that the proposed algorithms, in comparison to other algorithms, is capable of generating superior data placement and service deployment schemes to significantly reduce response latency.

Current state-of-the-art QoS prediction methods face two main limitations. Firstly, most existing QoS prediction approaches are centralized, gathering all user-service invocation QoS records for training and optimization, which causes privacy breaches. While some federated learning-based methods consider user privacy in a distributed way, they either directly upload local trained parameters or use simple encryption for global aggregation at the central server, thus failing to truly protect user privacy. Secondly, existing federated learning-based methods neglect distributed user-service topology and latent behavior-attribute correlations, compromising QoS prediction accuracy. To address these limitations, we propose a novel framework namedPrivacy-EnhancedFederated ExpandedGraphLearning (PE-FGL) for secure QoS prediction. It first conducts user-service expansion on the invocation graph with advanced privacy-preserving techniques, upgrading first-order local QoS invocations to high-order interaction relationships. Then, it extracts hybrid features from the expanded invocation graph via deep learning and graph residual learning. Finally, a two-layer secure mechanism of federated parameters aggregation is designed to enable collaborative learning among users through local parameter segmentation and global aggregation, achieving effective and secure QoS prediction. Extensive experiments on WS-DREAM demonstrate effective QoS prediction across multiple metrics while preserving privacy in user-service invocations.

Personal credit assessment, as an essential method to gauge an individual's credit status, exerts profound influences on personal economic life, risk control of financial institutions, establishment of social credit system, as well as the stableness of the overall economy. To address the challenges of delayed updates on bank information, malicious node attacks, issue of lacking user trust and data islands, this paper proposes a personal credit rating method based on asynchronous federated learning that is incentivized by copula reputation. This particular method creates a federated learning incentive mechanism based on copula variable dependence relationships that precisely assesses and measures the credibility of each distributed banking individual without acquiring original data, thereby preventing attacks from malicious nodes. Moreover, it puts forth a lightweight asynchronous federated learning mechanism according to each bank's local model's copula reputation, thereby optimizing the selection of participating nodes to reduce the system's overall cost, and settling the dilemma of bank information status update delays. Lastly, numerical results from real data sets indicate that the asynchronous federated learning personal credit assessment scheme based on copula reputation proposed has favorable accuracy, efficiency and enhanced security, efficiently safeguarding data privacy.

In the era of interconnected networks, digitization, and big data, Artificial Intelligence (AI) technology is developing rapidly and has been widely applied in various research and application fields through innovative patterns. This report first introduces the correlative research background, various service models, and key research issues in service computing. Then, it presents the latest research advancements and challenges of predicting Quality of Service (QoS) by deep learning techniques in different computing paradigms. Also, it provides the construction and application effectiveness of a Smart Recycling Collection Service Platform. Finally, it summarizes the future research hotspots and application trends in AI for service computing.

Composed image retrieval is designed to more accurately retrieve target images that align with user intentions by using a combination of reference images and descriptive modification texts. However, existing methods primarily focus on designing complex feature fusion networks while neglecting the prevalent issues of noise and inconsistent sample quality in training data, leading to insufficient cross-modal semantic alignment and sample relevance modeling. To address this, we propose an innovative Tripartite Alignment Network (TAN) that introduces a momentum distillation mechanism, leveraging the historical knowledge of a teacher network as additional super-vision to guide the optimization of the student network. During the feature encoder fine-tuning stage, we design response-based knowledge distillation and feature-based knowledge distillation techniques, explicitly strengthening modal alignment through composed-target contrastive learning and implicitly promoting modal fusion via composed-target matching learning. In the combiner training stage, we incorporate a lightweight combiner network and employ a cross-entropy-based matching loss function, encouraging high matching scores for relevant image-text pairs and low scores for irrelevant pairs. Extensive experiments on the FashionIQ and Shoes datasets demonstrate that TAN exhibits superior performance compared to existing state-of-the-art methods, with notable improvements in R@10 of +14.03% and +13.09%, respectively. These results affirm the effectiveness of momentum distillation in multimodal learning. Access the source code at https://github.com/Maserhe/TAN.

Recently, the edge data caching (EDC) problem has received much attention. It aims to appropriately cache data on edge servers. Existing EDC approaches suffer from a series of limitations. First, they often overlook the diverse characteristics of data, including caching costs and latency preferences. In reality, different types of data vary in size and require different storage resources for caching. The impact of specified latency preferences of edge users for different data on the quality of experience should be considered in the EDC problem. Second, the temporal dynamics of edge users’ data requests and distributions have been insufficiently addressed. To overcome these limitations systematically, this paper focuses on the problem of temporal-aware edge data caching with specified latency preference (TEDC). We first formulate the TEDC problem and transform it into an optimization problem with multiple objectives and global constraints and prove its ${\mathcal{N}}{\mathcal{P}}$-hardness. Then, we propose an optimal approach named TEDC-IP to solve this TEDC problem with the Integer Programming technique and a heuristic algorithm named TEDC-A for finding approximate solutions to large-scale TEDC problems efficiently. Extensive experiments are conducted on two widely-used real-world datasets to evaluate the performance of our approach. The results demonstrate that TEDC-IP and TEDC-A significantly outperform state-of-the-art approaches in finding approximate solutions in terms of the trade-off among multiple metrics.

With the surge in Web service users, user profiling has become increasingly prominent in personalized service recommender system. Graph Neural Networks (GNNs) has emerged as a key technology for user feature extraction. However, these methods mostly focus on modeling pairwise interaction relationships by type and overlook the high-order interaction relationships and deep semantic correlations. Moreover, GNNs’ limited receptive fields restrict their capacity to capture user high-order features effectively. To address these issues, we propose a novel framework for advanced user profiling named Heterogeneous Interaction Graph Transformer (HIGT). Firstly, HIGT constructs a weighted heterogeneous interaction graph from historical user-service interactions, using edge types for interaction modes and weights for their frequency. Secondly, it uses a Transformer to extract high-order semantic attribute correlations and enhance global understanding through self-attention, while proposing a structure-enhanced attention mechanism to incorporate the graph structure into the Transformer architecture for extracting high-order interaction features of users. This dual high-order feature learning method provides deeper insight into users’ preferences for Web services. Extensive experiments on two real-world e-commerce service datasets reveal that HIGT brings a significant performance boost compared with competing models for user profiling.

MOTIVATION: Gene regulatory networks (GRNs) encode gene regulation in living organisms, and have become a critical tool to understand complex biological processes. However, due to the dynamic and complex nature of gene regulation, inferring GRNs from scRNA-seq data is still a challenging task. Existing computational methods usually focus on the close connections between genes, and ignore the global structure and distal regulatory relationships. RESULTS: In this study, we develop a supervised deep learning framework, IGEGRNS, to infer GRNs from scRNA-seq data based on graph embedding. In the framework, contextual information of genes is captured by GraphSAGE, which aggregates gene features and neighborhood structures to generate low-dimensional embedding for genes. Then, the k most influential nodes in the whole graph are filtered through Top-k pooling. Finally, potential regulatory relationships between genes are predicted by stacking CNNs. Compared with nine competing supervised and unsupervised methods, our method achieves better performance on six time-series scRNA-seq datasets. AVAILABILITY AND IMPLEMENTATION: Our method IGEGRNS is implemented in Python using the Pytorch machine learning library, and it is freely available at https://github.com/DHUDBlab/IGEGRNS.

The proliferation of homogeneous web services has necessitated the task of predicting vacant Quality of Service (QoS) for service-oriented downstream tasks. Existing approaches primarily focus on user-service invocations without considering temporal factors, limiting their applicability in QoS fluctuations over time. Moreover, some investigations are conducted to predict temporally missing QoS, which still suffers from two limitations. First, time-aware collaborative filtering (CF) approaches fail to well capture continuous temporal changes, which lowers the performance of time-aware QoS prediction. Second, they have paid less attention to the high sparsity of user-service QoS invocations across sequentially multiple time slices, which affects the calculation of temporal average QoS, thereby further reducing the accuracy of time-aware QoS prediction. To effectively mine the continuous temporal variations and solve the high sparsity of user-service QoS invocations, we propose a novel time-aware QoS prediction approach named Temporal Reinforced Collaborative Filtering (TRCF). We design temporal reinforced RBS and PCC to improve similarity evaluation that leads to better calculation of temporal average QoS and deviation migration for predicting time-aware QoS. We evaluate TRCF on a large-scale real-world temporal dataset WS-DREAM across 64 time slices and the results demonstrate its superior performance in time-aware QoS prediction, both under relatively dense and extremely sparse QoS situations.

QoS prediction plays an important role in service-oriented downstream tasks. However, most of current state-of-the-art QoS prediction approaches suffer from two limitations. First, traditional approaches typically require collection of user-service historical QoS invocations centrally in order to improve QoS prediction accuracy, which poses a threat to user data privacy. Second, although few of the recent approaches take into account data protection when predicting QoS values, they still cannot effectively capture user-service complex nonlinear invocation relationships, significantly influencing the performance of QoS prediction. To address these two issues, we propose a novel framework of data-protected QoS prediction called Federated Residual Ladder Network (FRLN), which ensures user data protection and effectiveness of predicting missing QoS values. It initially leverages our designed Residual Ladder Network (RLN) to extract latent features of users and services from both low and high dimensional spaces. Then, local QoS prediction models are collaboratively trained by personalized federated learning with the consideration of data heterogeneity. Extensive experiments have been conducted on a real-world large-scale dataset called WS-DREAM, which consists of 5825 Web services from 74 regions and 339 users from 31 regions comprising a total number of 1,974,675 user-service QoS invocations. Experimental results demonstrate the effectiveness of FRLN in multiple evaluation metrics. While the proposed FRLN framework marks a significant step forward for QoS prediction in machine learning, ongoing advancements in ML techniques and expanded datasets are essential for further enhancing its precision and applicability in real-world scenarios.

In service recommender system, graph neural networks (GNNs) perform message passing through diffusion mechanism based on user-service relationship graph. However, existing GNN-based service recommendation models suffer from two limitations: ❨1❩ message passing is only carried out at firstorder neighbors, as higher-order may cause over-smoothing phenomenon, confining feature propagation in GNNs; and ❨2❩ due to sparse and noisy interactions, the distribution of embedding vectors is nonuniform in the latent space, resulting in unsatisfactory performance for downstream applications. To this end, we propose a fixed global graph diffusion view that is independent of the original user-service observed local view to form a multi-view learning by building contrastive learning (CL) relationship, named as Multi-View Graph Contrastive Learning (MVGCL). Specifically, it enhances the capability of message passing through constructed local and global multi-view graphs, and alleviates the sparse and noisy influences by performing intra-CL within local/global view and inter-CL between multi-view to obtain a more uniform distribution of user and service node representations. Extensive experiments are conducted on three benchmark datasets within different scales, and the results demonstrate that our proposed MVGCL can remarkably outperforms state-of-the-art competing baselines on various evaluation metrics.

Single-cell RNA sequencing (scRNA-seq) measures transcriptome-wide gene expression at single-cell resolution. Clustering analysis of scRNA-seq data enables researchers to characterize cell types and states, shedding new light on cell-to-cell heterogeneity in complex tissues. Recently, self-supervised contrastive learning has become a prominent technique for underlying feature representation learning. However, for the noisy, high-dimensional and sparse scRNA-seq data, existing methods still encounter difficulties in capturing the intrinsic patterns and structures of cells, and seldom utilize prior knowledge, resulting in clusters that mismatch with the real situation. To this end, we propose scDECL, a novel deep enhanced constraint clustering algorithm for scRNA-seq data analysis based on contrastive learning and pairwise constraints. Specifically, based on interpolated contrastive learning, a pre-training model is trained to learn the feature embedding, and then perform clustering according to the constructed enhanced pairwise constraint. In the pre-training stage, a mixup data augmentation strategy and interpolation loss is introduced to improve the diversity of the dataset and the robustness of the model. In the clustering stage, the prior information is converted into enhanced pairwise constraints to guide the clustering. To validate the performance of scDECL, we compare it with six state-of-the-art algorithms on six real scRNA-seq datasets. The experimental results demonstrate the proposed algorithm outperforms the six competing methods. In addition, the ablation studies on each module of the algorithm indicate that these modules are complementary to each other and effective in improving the performance of the proposed algorithm. Our method scDECL is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DBLABDHU/scDECL.

Drug combination therapy has gradually become a promising treatment strategy for complex or co-existing diseases. As drug-drug interactions (DDIs) may cause unexpected adverse drug reactions, DDI prediction is an important task in pharmacology and clinical applications. Recently, researchers have proposed several deep learning methods to predict DDIs. However, these methods mainly exploit the chemical or biological features of drugs, which is insufficient and limits the performances of DDI prediction. Here, we propose a new deep multimodal feature fusion framework for DDI prediction, DMFDDI, which fuses drug molecular graph, DDI network and the biochemical similarity features of drugs to predict DDIs. To fully extract drug molecular structure, we introduce an attention-gated graph neural network for capturing the global features of the molecular graph and the local features of each atom. A sparse graph convolution network is introduced to learn the topological structure information of the DDI network. In the multimodal feature fusion module, an attention mechanism is used to efficiently fuse different features. To validate the performance of DMFDDI, we compare it with 10 state-of-the-art methods. The comparison results demonstrate that DMFDDI achieves better performance in DDI prediction. Our method DMFDDI is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DHUDEBLab/DMFDDI.git.

MOTIVATION: Drug combination therapy has exhibited remarkable therapeutic efficacy and has gradually become a promising clinical treatment strategy of complex diseases such as cancers. As the related databases keep expanding, computational methods based on deep learning model have become powerful tools to predict synergistic drug combinations. However, predicting effective synergistic drug combinations is still a challenge due to the high complexity of drug combinations, the lack of biological interpretability, and the large discrepancy in the response of drug combinations in vivo and in vitro biological systems. RESULTS: Here, we propose DGSSynADR, a new deep learning method based on global structured features of drugs and targets for predicting synergistic anticancer drug combinations. DGSSynADR constructs a heterogeneous graph by integrating the drug-drug, drug-target, protein-protein interactions and multi-omics data, utilizes a low-rank global attention (LRGA) model to perform global weighted aggregation of graph nodes and learn the global structured features of drugs and targets, and then feeds the embedded features into a bilinear predictor to predict the synergy scores of drug combinations in different cancer cell lines. Specifically, LRGA network brings better model generalization ability, and effectively reduces the complexity of graph computation. The bilinear predictor facilitates the dimension transformation of the features and fuses the feature representation of the two drugs to improve the prediction performance. The loss function Smooth L1 effectively avoids gradient explosion, contributing to better model convergence. To validate the performance of DGSSynADR, we compare it with seven competitive methods. The comparison results demonstrate that DGSSynADR achieves better performance. Meanwhile, the prediction of DGSSynADR is validated by previous findings in case studies. Furthermore, detailed ablation studies indicate that the one-hot coding drug feature, LRGA model and bilinear predictor play a key role in improving the prediction performance. AVAILABILITY AND IMPLEMENTATION: DGSSynADR is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DHUDBlab/DGSSynADR.

Learning the similarity between fashion items is essential for many fashion-related tasks. Most methods based on global or local image similarity cannot meet the fine-grained retrieval requirements related to attributes. We are the first to clearly distinguish the concepts of attribute name and their values and divide fashion retrieval tasks that combine images and text into: attribute-guided retrieval and attribute-manipulated retrieval. We propose a hierarchical attribute-aware embedding network (HAEN) that takes images and attributes as input, learns multiple attribute-specific embedding spaces, and measures fine-grained similarity in the corresponding spaces. It can accurately map different attributes to the corresponding areas of the image, thereby facilitating the feature fusion of two different modalities of text and image, including enhancement and replacement. Then on this basis, we propose three attribute-manipulated similarity learning methods, HAEN_Avg, HAEN_Rec, and HAEN_Cmb. With comprehensive validation on two real-world fashion datasets, we demonstrate that our methods can effectively leverage semantic knowledge to improve image retrieval performance, including attribute-guided and attribute-manipulated retrieval tasks.

With the overwhelming explosion of Web services, how to effectively predict unknown QoS has become a key issue of differentiating large-scale similar or functionally equivalent Web services. However, current state-of-the-art QoS prediction approaches based on deep learning still suffer from two deficiencies. First, they mainly focus on predicting vacant QoS in a centralized manner and scarcely take into account distributed QoS prediction, which makes difficult to protect the privacy information of users invoking Web services. Second, they have ignored the hierarchical collaborative relationship to better extract latent features of users and services, reducing the accuracy of QoS prediction. To address these two issues, we propose a novel framework calledFederatedHierarchicalClustering forDistributedQoSPrediction(FHC-DQP). It collaboratively performs distributed federated training on independent users’ QoS invocations, and then the extracted federated users’ private features are fed to clustering algorithm for partitioning them into a set of clusters. By iteratively federated hierarchical clustering, users are fine-grained partitioned together and those users within the same cluster have stronger collaborative relevance for more effectively learning the latent features of users and services leading to the performance improvement of distributed QoS prediction, where contextual-aware deep neural network is designed for personalized QoS prediction. Extensive experiments are conducted based on a public real-world benchmarking dataset called WS-DREAM with almost 2,000,000 user-service historical QoS invocations. Compared with both centralized and federated competing baselines, the results demonstrate FHC-DQP receives superior performance for distributed QoS prediction, when it provides privacy-preserving of users’ QoS invocations.

Recently, edge user allocation (EUA) problem has received much attentions. It aims to appropriately allocate edge users to their nearby edge servers. Existing EUA approaches suffer from a series of limitations. First, considering users' service requests only as a whole, they neglect the fact that a service request may be partitioned into multiple tasks to be performed by different servers. Second, the impact of the spatial distance between edge users and servers on users' quality of experience is not properly considered. Third, the temporal dynamics of users' service requests has not been fully investigated. To overcome these limitations systematically, this paper focuses on the problem of spatio-temporal edge user allocation with task decomposition (ST-EUA). We first formulate the ST-EUA problem. Then, we transform ST-EUA problem as a multi-objective optimization problem and prove its NP-hardness. To tackle the ST-EUA problem effectively and efficiently, we propose a novel genetic algorithm-based heuristic approach GA-ST, aiming to maximize usersoverall QoE while minimizing migration cost in different time slots. Extensive experiments are conducted on two widely-used real-world datasets to evaluate the performance of GA-ST. The results demonstrate that GA-ST significantly outperforms state-of-the-art approaches in finding approximate solutions in terms of the trade-off among multiple metrics.

How to accurately predict vacant QoS has become a fundamental issue for service-oriented downstream tasks. However, most QoS prediction approaches based on model learning fail to discriminatively capture the latent feature representations of a user and a service, since they either leverage the shallow neural network such as MLP or take advantage of insufficient location information. Moreover, collaborative relationships of similar neighborhood have not been fully taken into account together with prediction model learning. To address these issues, we propose a novel framework for adaptive QoS prediction named Neighborhood-based Collaborative Residual Learning (NCRL). Location-aware two-tower deep residual network is designed to achieve neural QoS prediction by extracting latent features of users and services, which are fed to generate similar neighborhood for collaborative prediction based on historical QoS invocations. They are integrally combined to perform adaptive QoS prediction. Extensive experiments are conducted based on a large-scale real-world QoS dataset called WS-DREAM with almost 2,000,000 historical QoS invocations. The results indicate that NCRL can remarkably outperform state-of-the-art competing baselines.

The development of single-cell RNA-seq (scRNA-seq) technology allows researchers to characterize the cell types, states and transitions during dynamic biological processes at single-cell resolution. One of the critical tasks is to infer pseudo-time trajectory. However, the existence of transition cells in the intermediate state of complex biological processes poses a challenge for the trajectory inference. Here, we propose a new single-cell trajectory inference method based on transition entropy, named scTite, to identify transitional states and reconstruct cell trajectory from scRNA-seq data. Taking into account the continuity of cellular processes, we introduce a new metric called transition entropy to measure the uncertainty of a cell belonging to different cell clusters, and then identify cell states and transition cells. Specifically, we adopt different strategies to infer the trajectory for the identified cell states and transition cells, and combine them to obtain a detailed cell trajectory. For the identified cell clusters, we utilize the Wasserstein distance based on the probability distribution to calculate distance between clusters, and construct the minimum spanning tree. Meanwhile, we adopt the signaling entropy and partial correlation coefficient to determine transition paths, which contain a group of transition cells with the largest similarity. Then the transitional paths and the MST are combined to infer a refined cell trajectory. We apply scTite to four real scRNA-seq datasets and an integrated dataset, and conduct extensive performance comparison with nine existing trajectory inference methods. The experimental results demonstrate that the proposed method can reconstruct the cell trajectory more accurately than the compared algorithms. The scTite software package is available at https://github.com/dblab2022/scTite.

Single-cell RNA sequencing (scRNA-seq) permits researchers to study the complex mechanisms of cell heterogeneity and diversity. Unsupervised clustering is of central importance for the analysis of the scRNA-seq data, as it can be used to identify putative cell types. However, due to noise impacts, high dimensionality and pervasive dropout events, clustering analysis of scRNA-seq data remains a computational challenge. Here, we propose a new deep structural clustering method for scRNA-seq data, named scDSC, which integrate the structural information into deep clustering of single cells. The proposed scDSC consists of a Zero-Inflated Negative Binomial (ZINB) model-based autoencoder, a graph neural network (GNN) module and a mutual-supervised module. To learn the data representation from the sparse and zero-inflated scRNA-seq data, we add a ZINB model to the basic autoencoder. The GNN module is introduced to capture the structural information among cells. By joining the ZINB-based autoencoder with the GNN module, the model transfers the data representation learned by autoencoder to the corresponding GNN layer. Furthermore, we adopt a mutual supervised strategy to unify these two different deep neural architectures and to guide the clustering task. Extensive experimental results on six real scRNA-seq datasets demonstrate that scDSC outperforms state-of-the-art methods in terms of clustering accuracy and scalability. Our method scDSC is implemented in Python using the Pytorch machine-learning library, and it is freely available at https://github.com/DHUDBlab/scDSC.

Cellular programs often exhibit strong heterogeneity and asynchrony in the timing of program execution. Single-cell RNA-seq technology has provided an unprecedented opportunity for characterizing these cellular processes by simultaneously quantifying many parameters at single-cell resolution. Robust trajectory inference is a critical step in the analysis of dynamic temporal gene expression, which can shed light on the mechanisms of normal development and diseases. Here, we present TiC2D, a novel algorithm for cell trajectory inference from single-cell RNA-seq data, which adopts a consensus clustering strategy to precisely cluster cells. To evaluate the power of TiC2D, we compare it with three state-of-the-art methods on four independent single-cell RNA-seq datasets. The results show that TiC2D can accurately infer developmental trajectories from single-cell transcriptome. Furthermore, the reconstructed trajectories enable us to identify key genes involved in cell fate determination and to obtain new insights about their roles at different developmental stages.

As the 5th generation (5G) network develops and rolls out rapidly, user requests can be offloaded to nearby edge servers for processing. This alleviates the pressure on the network backhaul and the remote cloud. Nevertheless, the edge user allocation (EUA) problem, as one of the main research challenges in the 5G era, has become a major obstacle to ensuring users’ Quality of Experience (QoE) in the edge computing environment. Conventional EUA approaches, ranging from static global allocation model to online decision-making model, have ignored the long-term impact of the changes in users’ expectations on user-perceived Quality of Service (QoS). Additionally, most existing approaches have not taken into account the distance between an edge user and an edge server, which impacts the user’s data rate profoundly. In this paper, we tackle these challenges by formulating EUA problem as a spatial-temporal one (ST-EUA), which models distance-aware QoS based on the wireless transmission attenuation and models users’ QoE based on the Expectation Confirmation Theory (ECT). To find an appropriate solution for ST-EUA problem, we develop two fuzzy control-based approaches, namely FC and BFC, for on demand scenarios and batch processing scenarios, respectively. They can balance effectively the user consolidation and server load. We conduct extensive experiments based on two widely-used real-world datasets. The results demonstrate the superiority of our FC and BFC in effectiveness and efficiency over the baselines and state-of-the-art.

With the explosive growth in the number and diversity of Web services, correlative research has been investigated on Web service classification, as it fundamentally promotes advanced service-oriented applications, such as service discovery, selection, composition and recommendation. However, conventional approaches are restricted to indiscriminatingly classify Web services, which can trigger many challenges. First, they have not made full advantage of the implicit relationships among multi-dimensional information of Web services, such as the increasing number of service categories. Thus, it leads to low effectiveness of learning and representing service features, failing to ensure the overall accuracy of service classification. Second, the imbalance of service distributions has been ignored, while it is observed that service categories reveal distinct long-tail characteristics. That results in low accuracy on service classification for those categories that contain fewer Web services. To handle the challenges of more effectively learning implicit service features across the service repository, and with a particular concentration on those tail categories that contain fewer Web services, we propose a novel framework called DeepLTSC to more accurately perform the task of Web service classification under long-tail distributions. In DeepLTSC, we first present an improved label attentive convolutional deep neural network (LACNN) with service categories, which can generate deep service features to improve the overall classification performance. Then, a proposed service feature augmentation model (SFA) together with focal loss function is integrated into DeepLTSC to further optimize service features, aiming to boost the classification accuracy on tail service categories. Extensive experiments are conducted on three large-scale real-world services datasets with different long-tail distributions. The results demonstrate that DeepLTSC significantly outperforms state-of-the-art approaches for Web service classification on both overall and tail categories.

With an growing number of web services available on the Internet, an increasing burden is imposed on the use and management of service repository. Service clustering has been employed to facilitate a wide range of service-oriented tasks, such as service discovery, selection, composition and recommendation. Conventional approaches have been proposed to cluster web services by using explicit features, including syntactic features contained in service descriptions or semantic features extracted by probabilistic topic models. However, service implicit features are ignored and have yet to be properly explored and leveraged. To this end, we propose a novel heuristics-based framework DeepWSC for web service clustering. It integrates deep semantic features extracted from service descriptions by an improved recurrent convolutional neural network and service composability features obtained from service invocation relationships by a signed graph convolutional network, to jointly generate integrated implicit features for web service clustering. Extensive experiments are conducted on 8,459 real-world web services. The experiment results demonstrate that DeepWSC outperforms state-of-the-art approaches for web service clustering in terms of multiple evaluation metrics.

Nowadays, the world is witnessing a rapid development of edge computing. As an important issue in the edge computing paradigm, the edge user allocation (EUA) problem has attracted considerable attention. EUA aims at allocating the end-users in a specific area to the edge servers in that area, and ensure end-users' low-latency access to app vendor's services deployed on those edge servers. However, existing approaches simply assume that each edge server has a specific coverage and neglect the complexity of wireless signal transmission. To ensure end-users' low latency, an EUA approach must take into account the distance between end-users and their nearby edge servers, as it significantly impacts their Quality of Experience (QoE). Accordingly, EUA must maximize the overall QoE of the app vendor's users. To tackle this new distance-aware EUA problem, we propose two novel approaches, namely DEUA-O and DEUA-H. DEUA-O aims to find the optimal solution while DEUA-H aims to find the sub-optimal solution in large-scale scenarios efficiently. Four series of experiments are conducted on a real-world dataset to evaluate DEUA-O and DEUA-H. The results demonstrate the substantial gains of our approaches over the state-of-the-art.

Quality of service (QoS) has been mostly applied to represent non-functional properties of Web services and differentiate those with the same functionality. How to accurately predict service QoS has become a key research topic. Researchers have employed neighborhood information into matrix factorization (MF) for service QoS prediction in recent years. However, they are restricted to traditional matrix factorization that may incur a couple of limitations. 1) Conventional MF for QoS prediction linearly combines the multiplication of the latent feature representation of users and services through inner product, failing to fully capture the implicit features of user and service. 2) Most of approaches integrate user or service neighborhood as heuristics into MF model, where either location context or historical invocation records are used to calculate similar users or services. Nevertheless, combining both of them together in a collaborative way is ignored for neighborhood selection that has yet to be properly explored. To deal with the challenges, we propose a novel approach for service QoS prediction called Neighborhood-integrated Deep Matrix Factorization (NDMF), which integrates user neighborhood selected by a collaborative way into an enhanced matrix factorization model via deep neural network (DNN). We implement a prototype system and conduct extensive experiments on public and real-world large Web service dataset with almost 2,000,000 service invocations called WS-DREAM which is widely used in service QoS prediction. The experimental results demonstrate that our proposed approach significantly outperforms state-of-the-art ones in terms of multiple evaluation metrics.

With the development of cloud computing, the cloud market is becoming more and more complicated. There are many cloud providers and different cloud instance types, which brings users some confusion when they select cloud instance types. In order to solve the cloud instance type selection problem in multi-cloud environment, a Cloud Instance Type Selection Algorithm based on Genetic Algorithm (CITSA-GA) is proposed. CITSA-GA mainly includes two-dimensional encoding with the constraint between adjacent genes, selection operation adopting the elite retention strategy and the roulette strategy, crossover operation using the first fit strategy, and mutation operation with mutation bounds. We perform some experiments to prove the effectiveness of the proposed CITSA-GA.

Correlative approaches have attempted to cluster web services based on either the explicit information contained in service descriptions or functionality semantic features extracted by probabilistic topic models. However, the implicit contextual information of service descriptions is ignored and has yet to be properly explored and leveraged. To this end, we propose a novel framework with deep neural network, called DeepWSC, which combines the advantages of recurrent neural network and convolutional neural network to cluster web services through automatic feature extraction. The experimental results demonstrate that DeepWSC outperforms state-of-the-art approaches for web service clustering in terms of multiple evaluation metrics.

QoS-aware Web service composition has recently become one of the most challenging research issues. Although much work has been investigated to solve the problem, they mainly focus on certain QoS of Web services, while QoS with uncertainty exposes the most important characteristic in a real and highly dynamic environment on the Internet. In this paper, with the consideration of uncertain service QoS, we model the issue of Web service composition with QoS uncertainty that is translated into a multi-objective optimization problem via uncertain interval number, which can be solved by our proposed approach via an non-deterministic multi-objective evolutionary algorithm using the strategy of decomposition. Large-scale empirical experiments have been conducted on our simulated datasets. The experimental results demonstrate that our proposed approach can effectively and efficiently find an optimum composite service solution set with satisfactory convergence.

Multi-class classification problem is still a research hotspot in machine learning, and researchers dedicate themselves to create new algorithms with higher efficiency and accuracy. A Modularity-based Hierarchical Classification Tree (MHCT) is proposed in this paper, which derived from the idea of community detection, and the structure of the tree is similar to the hierarchical cluster tree. This classification approach is a supervised learning method combined with modularity for its convergence indicator. After the building process of the tree from bottom to top, several classifier predictors are created in the training step. Finally, the comparison experiments are conducted between our methods and other two kinds of popular multi-class classification algorithms (i.e. support vector machine and decision tree), using ten benchmark datasets from UCI (University of California, Irvine) machine learning repository. In this work, the experimental results indicate that the proposed methods have drastically reduced the excessive training time while maintaining accuracy is comparable to the other algorithms.

QoS values may significantly vary due to the invocation of Web services under dynamical network environment. Although some of approaches apply uncertain QoS to computing the skyline for the reduction of the number of candidate services, they have no uncertain QoS model that needs to be aligned to realistic invocation and execution of Web services. To solve the issue, this paper presents a novel approach to uncertain service skyline via Chebyshev's inequality and interval number. We model uncertain QoS of a Web service by a QoS matrix and then each dimension is shrank to an uncertain QoS scope by interval number. Finally, based on the QoS model, we propose an uncertain service skyline algorithm to compare the QoS of two Web services with domination relationship strategy. Extensive experiments have been conducted on 1,558,224 Web service invocation records. The experimental results demonstrate the effectiveness of our proposed approach.

BACKGROUND: Differentiation of human embryonic stem cells requires precise control of gene expression that depends on specific spatial and temporal epigenetic regulation. Recently available temporal epigenomic data derived from cellular differentiation processes provides an unprecedented opportunity for characterizing fundamental properties of epigenomic dynamics and revealing regulatory roles of epigenetic modifications. RESULTS: This paper presents a spatial temporal clustering approach, named STCluster, which exploits the temporal variation information of epigenomes to characterize dynamic epigenetic mode during cellular differentiation. This approach identifies significant spatial temporal patterns of epigenetic modifications along human embryonic stem cell differentiation and cluster regulatory sequences by their spatial temporal epigenetic patterns. CONCLUSIONS: The results show that this approach is effective in capturing epigenetic modification patterns associated with specific cell types. In addition, STCluster allows straightforward identification of coherent epigenetic modes in multiple cell types, indicating the ability in the establishment of the most conserved epigenetic signatures during cellular differentiation process.

Online social networks have found a significant increase in their popularity in recent years. All the networks have community structure, and one of the research problems mostly frequently tackled is the discovery of communities. An overlapping community is a network structure that allows one node to be a member of multiple communities. The method presented in this paper aims at detecting overlapping communities in social networks, and its novelty lies in that it combines with the Microblog User Model (MUM) which can reflect the interest of the user accurately. First, the MUM network, which is an undirected and weighted network, is constructed by computing the similarity among MUMs. Afterwords, Overlapping Community Detection based on MUM (OCD-MUM) is performed to partition the network. A community stops expanding when the fitness function reaches a local maximum. The communities detected are locally optimized. A user's interest is not only decided by the MUM, but it is also affected by the communities the user belongs to. The community model can reflect the interest of the community. The MUM is updated with community model of its communities, and therefore the interest of the user can be predicted by these communities. Our experiment result shows that OCD-MUM has a higher modularity Q value than traditional methods and the predicted interest is more close to the real world situations.

With the information overload in web services, micro-blog has been increasingly providing as a media for end-users to express their opinions. The notable feature of micro-blog articles is prone to be a burst of popularity during a short period. In addition, diverse interests make users bored in redundant items in most recommender systems. Therefore, providing users with diverse popular micro-blogs that suit their interesting topics is an important issue. In this paper, depending on forwarding number and comment number of micro-blogs, an effective model for popularity prediction is proposed to discover popular topics. Then, a MaxMin diversity algorithm based on content distance and popularity density is proposed to discover top k micro-blogs. Finally, we design a diverse personalized popularity attention (DPPA) recommendation approach for target user. We conduct extensive experiments on large scale micro-blog datasets. The experimental results show that our proposed approach can satisfy user's requirements with a higher recall than personal attention methods.

Nucleosomes are basic elements of chromatin structure. The positioning of nucleosomes along a genome is very important to dictate eukaryotic DNA compaction and access. Current computational methods have focused on the analysis of nucleosome occupancy and the positioning of well-positioned nucleosomes. However, fuzzy nucleosomes require more complex configurations and are more difficult to predict their positions. We analyzed the positioning of well-positioned and fuzzy nucleosomes from a novel structural perspective, and proposed WaveNuc, a computational approach for inferring their positions based on continuous wavelet transformation. The comparative analysis demonstrates that these two kinds of nucleosomes exhibit different propeller twist structural characteristics. Well-positioned nucleosomes tend to locate at sharp peaks of the propeller twist profile, whereas fuzzy nucleosomes correspond to broader peaks. The sharpness of these peaks shows that the propeller twist profile may contain nucleosome positioning information. Exploiting this knowledge, we applied WaveNuc to detect the two different kinds of peaks of the propeller twist profile along the genome. We compared the performance of our method with existing methods on real data sets. The results show that the proposed method can accurately resolve complex configurations of fuzzy nucleosomes, which leads to better performance of nucleosome positioning prediction on the whole genome.

Web service composition (WSC) is the task of combining a chain of connected single services together to create a more complex and value-added composite service. Quality of service (QoS) has been mostly applied to represent nonfunctional properties of web services and differentiate those with the same functionality. Many research has been done on QoS-aware service composition, as it significantly affects the quality of a composite service. However, existing methods are restricted to predefined workflows, which can incur a couple of limitations, including the lack of guarantee for the optimality on overall QoS and for the completeness of finding a composite service solution. In this paper, instead of predefining a workflow model for service composition, we propose a novel planning-based approach that can automatically convert a QoS-aware composition task to a planning problem with temporal and numerical features. Furthermore, we use state-of-the-art planners, including an existing one and a self-developed one, to handle complex temporal planning problems with logical reasoning and numerical optimization. Our approach can find a composite service graph with the optimal overall QoS value while satisfying multiple global QoS constraints. We implement a prototype system and conduct extensive experiments on large web service repositories. The experimental results show that our proposed approach largely outperforms existing ones in terms of solution quality and is efficient enough for practical deployment.

For Web service composition, choreography has recently received great attention and demonstrated a few key advantages over orchestration such as distributed control, fairness, data efficiency, and scalability. Automated design of choreography plans, especially distributed plans for multiple roles, is more complex and has not been studied before. Existing work requires manual generation assisted by model checking. In this paper, we propose a novel planning-based approach that can automatically convert a given composition task to a distributed choreography specification. Although planning has been used for orchestration, it is difficult to use planning for choreography, as it involves decentralized control, concurrent workflows, and contingency. We propose a few novel techniques, including compilation of contingencies, dependency graph analysis, and communication control, to handle these characteristics using planning. We theoretically show the correctness of this approach and empirically evaluate its practicability.