Real-world graphs or networks are usually heterogeneous, involving multiple types of nodes and relationships. Heterogeneous graph neural networks (HGNNs) can effectively handle these diverse nodes and edges, capturing heterogeneous information within the graph, thus exhibiting outstanding performance. However, most methods of HGNNs usually involve complex structural designs, leading to problems such as high memory usage, long inference time, and extensive consumption of computing resources. These limitations pose certain challenges for the practical application of HGNNs, especially for resource-constrained devices. To mitigate this issue, we propose the Spiking Heterogeneous Graph Attention Networks (SpikingHAN), which incorporates the brain-inspired and energy-saving properties of Spiking Neural Networks (SNNs) into heterogeneous graph learning to reduce the computing cost without compromising the performance. Specifically, SpikingHAN aggregates metapath-based neighbor information using a single-layer graph convolution with shared parameters. It then employs a semantic-level attention mechanism to capture the importance of different meta-paths and performs semantic aggregation. Finally, it encodes the heterogeneous information into a spike sequence through SNNs, simulating bioinformatic processing to derive a binarized 1-bit representation of the heterogeneous graph. Comprehensive experimental results from three real-world heterogeneous graph datasets show that SpikingHAN delivers competitive node classification performance. It achieves this with fewer parameters, quicker inference, reduced memory usage, and lower energy consumption.

We present a theoretical framework analyzing the relationship between data distributions and fairness guarantees in deep learning. Our work establishes novel bounds that explicitly account for data distribution heterogeneity across demographic groups, while introducing a formal analysis framework that minimizes expected loss differences across these groups. Moreover, we derive bounds for fairness errors and convergence rates, characterizing how distributional differences between groups affect the fundamental trade-off between fairness and accuracy. Through extensive experiments on diverse datasets across various modalities (image, tabular data, and text), including FairVision (eye disease detection), CheXpert (pleural effusion detection), HAM10000 (skin lesion classification), FairFace (facial attribute recognition), ACS Income (income prediction), CivilComments-WILDS (toxic comment detection), we validate our theoretical findings and demonstrate that differences in feature distributions across demographic groups significantly impact model fairness, with performance disparities particularly pronounced in racial categories. The theoretical bounds we derive corroborate these empirical observations, providing insights into the fundamental limits of achieving fairness in deep learning models when faced with heterogeneous data distributions. This work advances our understanding of fairness in AI and provides a theoretical foundation for developing more equitable algorithms. Motivated by these theoretical insights, particularly the link between feature distribution shifts and fairness gaps, we propose Fairness-Aware Regularization (FAR), a practical training objective that directly minimizes inter-group discrepancies in feature centroids and covariances to improve equitable performance. We validate the effectiveness of FAR across all datasets considered in this study, consistently observing improvements in overall AUC, ES-AUC, and subgroup performance.

Fairness remains a critical concern in healthcare, where unequal access to services and treatment outcomes can adversely affect patient health. While Federated Learning (FL) presents a collaborative and privacy-preserving approach to model training, ensuring fairness is challenging due to heterogeneous data across institutions, and current research primarily addresses non-medical applications. To fill this gap, we establish the first experimental benchmark for fairness in medical FL, evaluating six representative FL methods across diverse demographic attributes and imaging modalities. We introduce FairFedMed, the first medical FL dataset specifically designed to study group fairness (i.e., consistent performance across demographic groups). It comprises two parts: FairFedMed-Oph, featuring 2D fundus and 3D OCT ophthalmology samples with six demographic attributes; and FairFedMed-Chest, which simulates real cross-institutional FL using subsets of CheXpert and MIMIC-CXR. Together, they support both simulated and real-world FL across diverse medical modalities and demographic groups. Existing FL models often underperform on medical images and overlook fairness across demographic groups. To address this, we propose FairLoRA, a fairness-aware FL framework based on SVD-based low-rank approximation. It customizes singular value matrices per demographic group while sharing singular vectors, ensuring both fairness and efficiency. Experimental results on the FairFedMed dataset demonstrate that FairLoRA not only achieves state-of-the-art performance in medical image classification but also significantly improves fairness across diverse populations. Our code and dataset can be accessible via GitHub link: https://github.com/Harvard-AI-and-Robotics-Lab/FairFedMed.

ABSTRACT With the constant dynamics of temporal dependence and spatial correlation, the interaction between them has become intricate. Existing work attempts to model precise temporal dependency and spatial correlation to make their interactions more accurate but ignores the importance of understanding how the two interact with each other. Thus, this article mines deeper into their interaction mechanism and proposes a new traffic prediction model called traffic flow prediction model based on spatial–temporal identity (TPST). It provides a new way named the spatial–temporal identity mechanism to model spatial–temporal interactions, which convert complex temporal dependence and spatial correlation into their identity information. Meanwhile, in order to improve spatial–temporal interaction resolution of the model, the method utilizes the down‐sampling cross‐convolution technique to contain more spatial–temporal history information and parses spatial–temporal interactions at different granularity. Experiments conducted with four real traffic flow datasets show that TPST consistently outperforms the other seven benchmark models, providing higher prediction accuracy with lower computational cost.

Crime prediction, which focuses on forecasting the occurrence of criminal activities across city regions before they occur, constitutes an essential capability of surveillance systems designed to enhance urban security. While much effort has been invested in this field, most of the existing studies pay little attention to the influence of situational contexts on criminal activities, which hinders further improvement in prediction performance. To address this challenge, we propose a novel context-aware Bayesian tensor decomposition framework, namely cBTD-Crime, for citywide multi-step crime prediction. More specifically, cBTD-Crime first constructs a third-order tensor to simultaneously model spatial, temporal, and contextual factors and then applies the CP decomposition to exploit the intricate relationships between the three factors to facilitate the prediction process. To reduce the parameter tuning cost, cBTD-Crime further reformulates the problem from a probabilistic perspective, where a range of carefully selected distributions are placed on the spatial, temporal, and contextual latent factors. Finally, an efficient Gibbs sampling procedure is developed to generate a series of samples and the arithmetic mean is computed to obtain the predicted number of crime incidents. Experimental results show that cBTD-Crime achieves superior performance on real-world crime datasets in terms of different evaluation metrics.

The adoption of vehicular cloud computing for sequential recommendation offers flexible, reliable, and scalable computing resources in intelligent transportation systems. However, it also raises privacy concerns of drivers/passengers regarding the upload of sensitive data and models to vehicular cloud servers. To address this issue, we propose a novel privacy-preserving sequential recommendation method for intelligent transportation systems (named Grapeseed) based on split learning and variational autoencoder (VAE). Specifically, the vehicular client first inputs raw data into an encoder to produce latent variables locally and uploads these variables to the vehicular cloud server. Then, the vehicular cloud server generates and returns intermediate variables derived from these latent variables. Upon receiving these intermediate variables, the vehicular client calculates the final recommendation results. Extensive experiment results and analyses demonstrate that the proposed method improves both performance and communication efficiency between vehicular cloud servers and clients while preserving privacy.

Visible Light Positioning (VLP) has emerged as a promising technique in the Internet of Things landscape and gained increasing attention worldwide due to its widely existing infrastructure, high precision, and cost-effectiveness. Recently, ratio and difference-based VLP systems have been used to reduce errors from environmental noise, ambient light, and device differences. However, there may be intricate interference patterns that simple ratios and differences struggle to address. Moreover, a single LED often has limited capability to achieve self-diagnosis and self-correction. In fact, the information from other LEDs can be used to refine the signal and suppress interference. Thus, we propose to organize the VLP system in a graph and use the Graph Neural Network to model the interrelationships among LED lamps. This allows us to optimize the signals and further efficiently suppress interferences by simultaneously considering multiple LED lamps. In addition, the precisions of LEDs’ measurements is different due to various factors (e.g., distances and powers), and low-precision measurements may reduce the performance of the VLP system. To address this issue, we incorporate an attention layer to allow our model to give higher weights to high-precision measurements. Finally, the long short-term memory network is used to model the temporal dependencies between adjacent positions in a trajectory. Taking these modules together, we develop a robust VLP system called DeepVLP. The comprehensive experiments demonstrate that DeepVLP achieves better performance than state-of-the-art methods.

With the rapid development of cloud computing and service computing, service recommendation systems play a crucial role in helping users efficiently filter the appropriate services. However, the sparsity of service data and the presence of noise in interactions make it extremely challenging to accurately capture user preferences. Existing service recommendation methods based on Graph Neural Networks (GNNs) primarily rely on ID aggregation, often neglecting the richness of textual semantics and are susceptible to interaction noise, resulting in suboptimal modeling of user-service relationships. Although Large Language Models (LLMs) demonstrate remarkable advantages in capturing textual semantics, current methods struggle to effectively align structural representations with textual representations, limiting improvements in recommendation performance. To address these challenges, we propose an innovative multi-semantic graph diffusion model for service recommendation, DiffMSR, which aims to align textual and structural representations while learning the generation process of interaction graphs in a denoising manner. This approach mitigates data sparsity and effectively reduces noise interference. Specifically, the model leverages LLMs to capture the textual semantic features of service descriptions and integrates them with structured semantic information from knowledge graphs. Through cross-semantic contrastive learning, it achieves heterogeneous semantic alignment. Furthermore, the model introduces a multisemantic diffusion-based generation framework, which iteratively denoises to construct high-quality user-service interaction graphs. This significantly enhances the multi-semantic awareness of user representations, thereby improving recommendation performance. Experiments on public service datasets demonstrate that DiffMSR outperforms existing state-of-the-art baseline methods, achieving improvements of 4.13% and 6.37% in recommendation accuracy and recall, respectively.

Fairness is a critical concern in deep learning, especially in healthcare, where these models influence diagnoses and treatment decisions. Although fairness has been investigated in the vision-only domain, the fairness of medical vision-language (VL) models remains unexplored due to the scarcity of medical VL datasets for studying fairness. To bridge this research gap, we introduce the first fair vision-language medical dataset (Harvard-FairVLMed) that provides detailed demographic attributes, ground-truth labels, and clinical notes to facilitate an in-depth examination of fairness within VL foundation models. Using Harvard-FairVLMed, we conduct a comprehensive fairness analysis of two widely-used VL models (CLIP and BLIP2), pre-trained on both natural and medical domains, across four different protected attributes. Our results highlight significant biases in all VL models, with Asian, Male, Non-Hispanic, and Spanish being the preferred subgroups across the protected attributes of race, gender, ethnicity, and language, respectively. In order to alleviate these biases, we propose FairCLIP an optimal-transport-based approach that achieves a favorable trade-off between performance and fairness by reducing the Sinkhorn distance between the overall sample distribution and the distributions corresponding to each demographic group. As the first VL dataset of its kind, Harvard-FairVLMed holds the potential to catalyze advancements in the development of machine learning models that are both ethically aware and clinically effective. Our dataset and code are available at https://ophai.hms.harvard.edu/datasets/harvard-fairvlmed10k.

Fairness in artificial intelligence models has gained significantly more attention in recent years, especially in the area of medicine, as fairness in medical models is critical to people's well-being and lives. High-quality medical fairness datasets are needed to promote fairness learning research. Existing medical fairness datasets are all for classification tasks, and no fairness datasets are available for medical segmentation, while medical segmentation is an equally important clinical task as classifications, which can provide detailed spatial information on organ abnormalities ready to be assessed by clinicians. In this paper, we propose the first fairness dataset for medical segmentation named Harvard-FairSeg with 10,000 subject samples. In addition, we propose a fair error-bound scaling approach to reweight the loss function with the upper error-bound in each identity group, using the segment anything model (SAM). We anticipate that the segmentation performance equity can be improved by explicitly tackling the hard cases with high training errors in each identity group. To facilitate fair comparisons, we utilize a novel equity-scaled segmentation performance metric to compare segmentation metrics in the context of fairness, such as the equity-scaled Dice coefficient. Through comprehensive experiments, we demonstrate that our fair error-bound scaling approach either has superior or comparable fairness performance to the state-of-the-art fairness learning models. The dataset and code are publicly accessible via https://ophai.hms.harvard.edu/datasets/harvard-fairseg10k.

Visible Light Positioning (VLP), a promising indoor positioning technique, has gained wide popularity worldwide because of its ubiquitous infrastructure, low power consumption, and high positioning precision. However, VLP systems based on photodiodes (PDs) often suffer from varying ambient light with time and space, which seriously degrades their positioning precision and robustness. In this article, we carefully evaluate the influence of the ambient light on the VLP system, which includes the reduction of positioning accuracy by varying ambient light with time and the inaccurate parameter calibration by unevenly distributed ambient light. Then, we figure out that the influence of ambient light on the Received Signals Strength (RSS) values is determined by the ambient light intensity and PD, which is independent of external factors, including distance, frequency, LED, etc. Next, we propose a new positioning framework, RatioVLP, where a ratio model that is more robust to varying ambient light with time is used. However, the ratio model is severely dependent on the Lambert parameters that are vulnerable to ambient light, which reduces the framework's precision when the calibration area is unevenly covered by ambient light. Thus, we design new parameters that are less sensitive to ambient light, calledR parameter, to connect the RSS ratio and its corresponding distance ratio, which can strengthen the ratio model's robustness and effectively reduce the influence of ambient light on the parameter calibration process. Experimental results show that the positioning precision of the proposed method is improved by more than 50 % when compared to the conventional Lambert model in scenes influenced by ambient light.

Fairness (also known as equity interchangeably) in machine learning is important for societal well-being, but limited public datasets hinder its progress. Currently, no dedicated public medical datasets with imaging data for fairness learning are available, though underrepresented groups suffer from more health issues. To address this gap, we introduce Harvard Glaucoma Fairness (Harvard-GF), a retinal nerve disease dataset including 3,300 subjects with both 2D and 3D imaging data and balanced racial groups for glaucoma detection. Glaucoma is the leading cause of irreversible blindness globally with Blacks having doubled glaucoma prevalence than other races. We also propose a fair identity normalization (FIN) approach to equalize the feature importance between different identity groups. Our FIN approach is compared with various state-of-the-art fairness learning methods with superior performance in the racial, gender, and ethnicity fairness tasks with 2D and 3D imaging data, demonstrating the utilities of our dataset Harvard-GF for fairness learning. To facilitate fairness comparisons between different models, we propose an equity-scaled performance measure, which can be flexibly used to compare all kinds of performance metrics in the context of fairness. The dataset and code are publicly accessible via https://ophai.hms.harvard.edu/datasets/harvard-gf3300/.

The SPiForest, a new isolation-based approach to outlier detection, constructs iTrees on the space containing all attributes by probability density-based inverse sampling. Most existing iForest (iF)-based approaches can precisely and quickly detect outliers scattering around one or more normal clusters. However, the performance of these methods seriously decreases when facing outliers whose nature "few and different" disappears in subspace (e.g., anomalies surrounded by normal samples). To solve this problem, SPiForest is proposed, which is different from existing approaches. First, SPiForest uses the principal component analysis (PCA) to find principal components and estimate each component's probability density function (pdf). Second, SPiForest utilizes the inv-pdf, which is inversely proportional to the pdf estimated from the given dataset, to generate support points in the space containing all attributes. Third, the hyperplane decided by these support points is used to isolate the outliers in the space. Next, these steps are repeated to build an iTree. Finally, many iTrees construct a forest for outlier detection. SPiForest provides two benefits: 1) it isolates outliers with fewer hyperplanes, which significantly improves the accuracy and 2) it effectively detects the outliers whose nature "few and different" disappears in subspace. Comparative analyses and experiments show that the SPiForest achieves a significant improvement in terms of area under the curve (AUC) when compared with the state-of-the-art methods. Specifically, our method improves by at most 17.7% on AUC when compared to iF-based algorithms.

Glaucoma is the number one cause of irreversible blindness globally. A major challenge for accurate glaucoma detection and progression forecasting is the bottleneck of limited labeled patients with the state-of-the-art (SOTA) 3D retinal imaging data of optical coherence tomography (OCT). To address the data scarcity issue, this paper proposes two solutions. First, we develop a novel generalization-reinforced semi-supervised learning (SSL) model called pseudo supervisor to optimally utilize unlabeled data. Compared with SOTA models, the proposed pseudo supervisor optimizes the policy of predicting pseudo labels with unlabeled samples to improve empirical generalization. Our pseudo supervisor model is evaluated with two clinical tasks consisting of glaucoma detection and progression forecasting. The progression forecasting task is evaluated both unimodally and multimodally. Our pseudo supervisor model demonstrates superior performance than SOTA SSL comparison models. Moreover, our model also achieves the best results on the publicly available LAG fundus dataset. Second, we introduce the Harvard Glaucoma Detection and Progression (Harvard-GDP) Dataset, a multimodal multitask dataset that includes data from 1,000 patients with OCT imaging data, as well as labels for glaucoma detection and progression. This is the largest glaucoma detection dataset with 3D OCT imaging data and the first glaucoma progression forecasting dataset that is publicly available. Detailed sex and racial analysis are provided, which can be used by interested researchers for fairness learning studies. Our released dataset is benchmarked with several SOTA supervised CNN and transformer deep learning models. The dataset and code are made publicly available via https://ophai.hms.harvard.edu/datasets/harvard-gdp1000.

As a critical topic of Internet of Things applications, smartphone-based indoor navigation has a rapidly growing need in various applications. However, indoor navigation technology is unreliable when facing a challenge in complex indoor environments. This article presents a tightly coupled (TC) integration of pedestrian dead reckoning (PDR) and Bluetooth for indoor pedestrian navigation and enhances it from three approaches. We first establish a Gaussian-based distance model (GDM) that improves the signal path-loss model to incorporate the prior information on the variation of signal volatility with distance. Then, the use of map information and a back-off strategy to optimize the particle transfer strategy further improves the positioning accuracy and rationality of the system. Moreover, we leverage behavioral landmarks, signal landmarks, and distance information to build a graph optimization model to optimize the proposed navigator. We have extensively verified the proposed navigator and compared it with the existing solutions and systems. Experimental results demonstrated that the average errors of the proposed solutions in three scenes were 34.71% of Bluetooth, 14.04% of PDR, 45.13% of the extended Kalman filter, 57.83% of the unscented Kalman filter, and 56.10% of PF, respectively. The results showed that our proposed solution has apparent advantages, especially when addressing the issues of incorrect trajectory updating and divergence of the system in a complex environment.

Ophthalmic images, along with their derivatives like retinal nerve fiber layer (RNFL) thickness maps, play a crucial role in detecting and monitoring eye diseases such as glaucoma. For computer-aided diagnosis of eye diseases, the key technique is to automatically extract meaningful features from ophthalmic images that can reveal the biomarkers (e.g., RNFL thinning patterns) associated with functional vision loss. However, representation learning from ophthalmic images that links structural retinal damage with human vision loss is non-trivial mostly due to large anatomical variations between patients. This challenge is further amplified by the presence of image artifacts, commonly resulting from image acquisition and automated segmentation issues. In this paper, we present an artifact-tolerant unsupervised learning framework called EyeLearn for learning ophthalmic image representations in glaucoma cases. EyeLearn includes an artifact correction module to learn representations that optimally predict artifact-free images. In addition, EyeLearn adopts a clustering-guided contrastive learning strategy to explicitly capture the affinities within and between images. During training, images are dynamically organized into clusters to form contrastive samples, which encourage learning similar or dissimilar representations for images in the same or different clusters, respectively. To evaluate EyeLearn, we use the learned representations for visual field prediction and glaucoma detection with a real-world dataset of glaucoma patient ophthalmic images. Extensive experiments and comparisons with state-of-the-art methods confirm the effectiveness of EyeLearn in learning optimal feature representations from ophthalmic images.

Knowledge representation of networked systems is fundamental in many disciplines. To date, existing methods for representation learning primarily focus on networks with simplex labels, yet real-world objects (nodes) are inherently complex in nature and often contain rich semantics or labels. For example, a user may belong to diverse interest groups of a social network, resulting in multi-label networks for many applications. A multi-label network not only has multiple labels for each node, the labels are often highly correlated making existing methods ineffective or even fail to handle such correlation for node representation learning. In this article, we propose a novel multi-label graph convolutional network (MuLGCN) for learning node representation. To fully explore label-label correlation and network topology structures, we propose to model a multi-label network as two Siamese GCNs: a node-node-label graph and a label-label-node graph. The two GCNs each handle one aspect of representation learning for nodes and labels, respectively, and are seamlessly integrated in one objective function. The learned label representations can effectively preserve the intra-label interaction and node label properties, and are aggregated to enhance the node representation learning under a unified training framework. Experiments and comparisons on multi-label node classification validate the effectiveness of our proposed approach.

Noise and inconsistency commonly exist in real-world information networks, due to the inherent error-prone nature of human or user privacy concerns. To date, tremendous efforts have been made to advance feature learning from networks, including the most recent graph convolutional networks (GCNs) or attention GCN, by integrating node content and topology structures. However, all existing methods consider networks as error-free sources and treat feature content in each node as independent and equally important to model node relations. Noisy node content, combined with sparse features, provides essential challenges for existing methods to be used in real-world noisy networks. In this article, we propose feature-based attention GCN (FA-GCN), a feature-attention graph convolution learning framework, to handle networks with noisy and sparse node content. To tackle noise and sparse content in each node, FA-GCN first employs a long short-term memory (LSTM) network to learn dense representation for each node feature. To model interactions between neighboring nodes, a feature-attention mechanism is introduced to allow neighboring nodes to learn and vary feature importance, with respect to their connections. By using a spectral-based graph convolution aggregation process, each node is allowed to concentrate more on the most determining neighborhood features aligned with the corresponding learning task. Experiments and validations, w.r.t. different noise levels, demonstrate that FA-GCN achieves better performance than the state-of-the-art methods in both noise-free and noisy network environments.

In traditional graph neural networks (GNNs), graph convolutional learning is carried out through topology-driven recursive node content aggregation for network representation learning. In reality, network topology and node content each provide unique and important information, and they are not always consistent because of noise, irrelevance, or missing links between nodes. A pure topology-driven feature aggregation approach between unaligned neighborhoods may deteriorate learning from nodes with poor structure-content consistency, due to the propagation of incorrect messages over the whole network. Alternatively, in this brief, we advocate a co-alignment graph convolutional learning (CoGL) paradigm, by aligning topology and content networks to maximize consistency. Our theme is to enforce the learning from the topology network to be consistent with the content network while simultaneously optimizing the content network to comply with the topology for optimized representation learning. Given a network, CoGL first reconstructs a content network from node features then co-aligns the content network and the original network through a unified optimization goal with: 1) minimized content loss; 2) minimized classification loss; and 3) minimized adversarial loss. Experiments on six benchmarks demonstrate that CoGL achieves comparable and even better performance compared with existing state-of-the-art GNN models.

Clustering Web services can improve the quality and efficiency of service discovery and management within a service repository. Nowadays, Web services frequently interact (e.g., composition relation and tag sharing relation) with each other to form a complex and heterogeneous service relationship network. The rich network relations inherently reflect either positive or negative clustering association between Web services, which can be a strong supplement to service semantics for characterizing functional affinities between Web services. In this paper, we propose to cluster Web services by utilizing both description documents and the structural information from the service relationship network. We first learn the content semantic information from service description documents based on the widely used Doc2vec model, and meanwhile, learn the structural semantic information from the service relationship network based on a network representation learning algorithm. Then, we propose to pretrain the content and structural semantic information to obtain the most relevant and unified features through training a service classification model with partially labeled data. Finally, a spectral clustering algorithm is utilized for Web services clustering based on the above unified features with preserved content and structural semantics. Therefore, the proposed services clustering approach takes advantage of both service content semantic and service network structure semantic based similarity between services. Extensive experiments are conducted on a real-world dataset from ProgrammableWeb, composed of 12919 Web API services. Experimental results demonstrate that our approach yields an improvement of 4.78% in precision and 5.4% in recall over the state-of-the-art method.

Extensive efforts have been applied to develop efficient feature extraction algorithms, which aim to achieve optimal results in many fundamental tasks such as Web-based software service clustering, recommendation and composition. However, one common issue for existing methods is that mined features are problem dependent, causing poor generalization ability across different applications. Recent studies show that we can represent networked data (e.g., citation networks and social networks) as low-dimensional vectors with rich structure and content information preserved, which can then greatly facilitate many downstream tasks such as classification and clustering. In this article, we focus on the problem of Web service network embedding, which aims to learn low-dimensional vectors to represent services by encoding both Mashup-API composition structure and service functional content. We first propose a novel probabilistic topic model to predict potential links between Mashups and APIs in the service network. Then, we develop a Service Graph Convolutional Network (Service-GCN) to learn vector representations of services, where each service (e.g., Mashup or API) forms its representation through message passing between neighborhood services over the network. We evaluate the network embedding quality on two real-world datasets for downstream classification and clustering tasks. Experimental results show that the average performance of our method improves 20.7 percent (Micro-F1) in service classification and 19.0 percent (Accuracy) in Mashup clustering compared to the state-of-the-art, which verified the effectiveness of the proposed approach for learning vector representations of Web services.

State estimation problems without absolute position measurements routinely arise in navigation of unmanned aerial vehicles, autonomous ground vehicles, etc., whose proper operation relies on accurate state estimates and reliable covariances. Unaware of absolute positions, these problems have immanent unobservable directions. Traditional causal estimators, however, usually gain spurious information on the unobservable directions, leading to over-confident covariance inconsistent with actual estimator errors. The consistency problem of fixed-lag smoothers (FLSs) has only been attacked by the first estimate Jacobian (FEJ) technique because of the complexity to analyze their observability property. But the FEJ has several drawbacks hampering its wide adoption. To ensure the consistency of a FLS, this paper introduces the right invariant error formulation into the FLS framework. To our knowledge, we are the first to analyze the observability of a FLS with the right invariant error. Our main contributions are twofold. As the first novelty, to bypass the complexity of analysis with the classic observability matrix, we show that observability analysis of FLSs can be done equivalently on the linearized system. Second, we prove that the inconsistency issue in the traditional FLS can be elegantly solved by the right invariant error formulation without artificially correcting Jacobians. By applying the proposed FLS to the monocular visual inertial simultaneous localization and mapping (SLAM) problem, we confirm that the method consistently estimates covariance similarly to a batch smoother in simulation and that our method achieved comparable accuracy as traditional FLSs on real data.

Real-world networked systems often show dynamic properties with continuously evolving network nodes and topology over time. When learning from dynamic networks, it is beneficial to correlate all temporal networks to fully capture the similarity/relevance between nodes. Recent work for dynamic network representation learning typically trains each single network independently and imposes relevance regularization on the network learning at different time steps. Such a snapshot scheme fails to leverage topology similarity between temporal networks for progressive training. In addition to the static node relationships within each network, nodes could show similar variation patterns (e.g., change of local structures) within the temporal network sequence. Both static node structures and temporal variation patterns can be combined to better characterize node affinities for unified embedding learning. In this paper, we propose Graph Attention Evolving Networks (GAEN) for dynamic network embedding with preserved similarities between nodes derived from their temporal variation patterns. Instead of training graph attention weights for each network independently, we allow model weights to share and evolve across all temporal networks based on their respective topology discrepancies. Experiments and validations, on four real-world dynamic graphs, demonstrate that GAEN outperforms the state-of-the-art in both link prediction and node classification tasks.

Summary With the increasing number of Web services, Web service discovery for service‐oriented application development has become more important. Clustering or classifying Web services according to their functionalities is an effective way for Web service discovery. Extracting latent topic features from service description by exploiting topic model can improve the accuracy of service classification. However, most of them simply treat the description document as a set of flat word features without considering the varying importance of different features as well as sequential relations between features. In this article, we proposed a Web service classification approach based on information gain theory and bidirectional long short‐term memory with attention mechanism for accuracy Web service classification by considering fine‐grained factors implicit in Web service description. The comparative experiments are performed on ProgrammableWeb dataset, and show that the proposed method achieves a significant improvement compared with baseline methods.

Tagging systems have been widely used as a major way of managing Web service resources. Many portals such as ProgrammableWeb and BioCatalogue allow users to create manual tags annotating Web services and their compositions (e.g., mashups). This is extremely helpful for managing and retrieving enormous Web service data. In the past few years, many tag recommendation approaches have been proposed for Web services that contain few or no tags. Most of them only exploit the textual content or tag service matrix information. Sometimes those approaches suffer from the data sparsity problem, especially when Web services have only few tags or their auxiliary textual contents are hard to be obtained. In real world, a plenty of relationships are available in recommendation systems, e.g., the composition relationship between services and the annotation relationship between mashups and tags. These multi-relational data can be utilized as additional features to improve the recommendation performance. In this paper, we exploit various types of relationships as features and propose a novel topic-sensitive approach based on the Factorization Machines for mashup tag recommendation. Factorization Machines is utilized to model the pair-wise interactions between all features and predict adequate tags for mashups. In this approach, we first obtain the latent topics of all tags as well as the description documents for mashups and APIs based on a novel probabilistic topic model. Then, a multi-relational network by mining various relationships from the Web service data is constructed. Various auxiliary informations are subsequently extracted from the network to train the Factorization Machines. The proposed model is evaluated on three real-world datasets and the experimental results show that it outperforms several state-of-the-art methods.

With the explosive growth of Web APIs on the Internet, it is a challenge to recommend desirable Web APIs from multiple ecosystems to develop a Mashup. Most existing API service recommendation methods focus on functional semantic similarity, but underutilize the rich network relations which inherently reflect either positive or negative relevance between services. Moreover, in the recommendation process, they usually pay too much attention to the interactions between Mashups and APIs, but ignore the cooperation between APIs. In this paper, we propose a novel method named SRAWN (Structure Reinforcing and Attribute Weakening Network) based API recommendation approach for Mashup creation. Specifically, we first design a feature extractor layer to capture structure relationship and attribute information from an API relation network graph by introducing a GAT2VEC framework, and obtain representation vectors corresponding to each API. Then, a matching evolving layer is proposed to capture the matching evolving process between APIs. At this layer, APIs are chosen incrementally to composite a Mashup, and the embedding vectors of the Mashup's existing composition features are updated adaptively based on diverse candidate APIs, by introducing a Deep Interest Network. Comprehensive experiments on a real-world dataset show that SRAWN outperforms the other state-of-the-art solutions.

Networked data often demonstrate the Pareto principle (i.e., 80/20 rule) with skewed class distributions, where most vertices belong to a few majority classes and minority classes only contain a handful of instances. When presented with imbalanced class distributions, existing graph embedding learning tends to bias to nodes from majority classes, leaving nodes from minority classes under-trained. In this paper, we propose Dual-Regularized Graph Convolutional Networks (DR-GCN) to handle multi-class imbalanced graphs, where two types of regularization are imposed to tackle class imbalanced representation learning. To ensure that all classes are equally represented, we propose a class-conditioned adversarial training process to facilitate the separation of labeled nodes. Meanwhile, to maintain training equilibrium (i.e., retaining quality of fit across all classes), we force unlabeled nodes to follow a similar latent distribution to the labeled nodes by minimizing their difference in the embedding space. Experiments on real-world imbalanced graphs demonstrate that DR-GCN outperforms the state-of-the-art methods in node classification, graph clustering, and visualization.

Network embedding aims to preserve topological structures of a network using low-dimensional vectors and has shown to be effective for driving a myriad of graph mining tasks (e.g., link prediction or classification) free of the stressful feature extraction procedure. Many methods have been proposed to integrate node content and/or label information, with nodes sharing similar content/labels being close to each other in the learned latent space. To date, existing methods either consider networked instances with a single label or consider a set of labels as a whole for node representation learning. Therefore, they cannot handle network of instances containing multiple labels (i.e. multi-labels), which are ubiquitous in describing complex concepts of instances. In this article, we formulate a new multi-label network embedding (MLNE) problem to learn feature representation for networked multi-label instances. We argue that the key to MLNE learning is to aggregate node topology structures, node content, and multi-label correlations. We propose a two-layer network embedding framework to couple information for effective learning. To capture higher order label correlations, we use labels to form a high-level label-label network over a low-level node-node network, in which the label network interacts with the node network through multi-labeling relations. The low-level node-node network can be enhanced by latent label-specific features from high-level label network with well-captured high-order correlations between labels. To enable the multi-label informed network embedding, we force both node and label representations being optimized under the same low-dimensional latent space by a unified training objective. Experiments on real-world data sets demonstrate that MLNE achieves better performance compared with methods with or without considering label information.

The advent of Service-Oriented Architecture (SOA) has brought a fundamental shift in the way in which distributed applications are implemented. An overwhelming number of Web-based services (e.g., APIs and Mashups) have leveraged this shift and furthered development. Applications designed with SOA principles are typically characterized by frequent dependencies with one another in the form of heterogeneous networks, i.e., annotation relations between tags and services, and composition relations between Mashups and APIs. Although prior work has shown the utility gained by exploring these networks, their analysis is still in its infancy. This article develops an approach to learning representations of the Web service network, which seeks to embed Web services in low-dimensional continuous vectors with preserved information of the network structure, functional tags, and service descriptions, such that services with similar functional properties and network structures are mapped together in the learned latent space. We first propose a topic generative model for constructing two topic distribution networks (Mashup-Topic and API-Topic) from the service content. Then, we present an efficient optimization process to derive low-dimensional vector representations of Web services from a tri-layer bipartite network with the Mashup-Topic and API-Topic networks on two ends and the Mashup-API composition network in the middle. Experiments on real-word datasets have verified that our approach is effective to learn robust low-rank service representations, i.e., 25% F1-measure gain over the state-of-the-art in Web service recommendation task.

Clustering Web services can promote the quality of services discovery and management within a service repository. Traditional clustering methods primarily focus on using the semantic distance between service features, i.e., latent topics learned from WSDL documents, to measure the service content similarity between Web services. Few works exploited the structural information generated during the usage of Web services, i.e., the service compositing and tagging behaviors. Nowadays, Web services frequently interact (e.g., composition relation and tag sharing relation) with each other to form a complex service relationship network. The rich network relations inherently reflect either positive or negative categorical relevance between services, which can be strong supplement of service semantics in characterizing the functional affinities between services. In this paper, we propose to utilize the services relationship network for augmented services clustering algorithm design. We first learn semantic information from service descriptions based on the widely used Doc2Vec model. Then, we propose a revised K-means algorithm for service clustering that benefits simultaneously from service semantics and network relations, where the service relations are previously preserved in a set of low-dimensional vectors achieved based on a recently proposed network embedding technique. Experiments on a real-world dataset demonstrated that the proposed clustering approach yields an improvement of 6.89% than the state-of-the-art.

The service-oriented architecture makes it possible for developers to create value-added Mashup applications by composing multiple available Web services. Due to the overwhelming number of Web services online, it is often hard and time-consuming for developers to find their desired ones from the entire service repository. In the past, various approaches aim at recommending Web services for automatic Mashup creation have been proposed, i.e., TFIDF, collaborative filtering and topic model-based methods, which rely on the original service descriptions given by service providers. However, most traditional methods fail to capture the function-related features of services since words contained in service descriptions usually correspond to different intent aspects (e.g., functional and non-functional related). To tackle this problem, we propose a tag attention-based recurrent neural networks model for Web service recommendation. The model consists of two Siamese bidirectional Long Short-Term Memory (LSTM) networks, which jointly learn two embeddings representing the functional features of Web services and the functional requirements of Mashups. In addition, by considering the tags of services as functional context information, the model can learn to assign attention scores to different words in service descriptions according to their intent importance, thus words used to reveal the functional properties of Web service will be given special attention. We compare our approach with the state-of-the-art methods (e.g., RTM, Word2vec, etc.) on a real-world dataset crawled from ProgrammableWeb, and the experimental results demonstrate the effectiveness of the proposed model.

How to provide a low-cost but accurate localization solution for the indoor mobile robots are essential in many Internet of Things applications, such as smart home and asset tracking. To achieve this goal, this paper originally proposes a modified two-filter smoother based on ensemble Kalman filter (KF) (denoted as EnKS) for the localization of indoor mobile robots. The proposed EnKS algorithm consists of both a forward part of an ensemble KF (EnKF) with statistical linear regression and a backward part of a modified information KF with state error vector. The EnKS based on stochastic sampling with ensemble members can achieve better positioning accuracy than other Kalman smoothers. When compared to EnKF, the proposed EnKS combines a backward filter to compensate for the estimation error of EnKF and further improves the accuracy. Furthermore, the implementation of the proposed EnKS is conducted in the real world visible light positioning (VLP) system using pre-existing LED lights for low-cost robot localization. To make a performance comparison, this paper also uses baseline smoothers based on extended KF and central difference KF in the VLP system. Preliminary experimental results imply that the proposed EnKS is able to achieve the best positioning accuracy, as high as 11.18 cm on average, but with a comparable computational complexity, which enables to meet the demands of many robot applications.

Service recommendation is a fundamental task in many application environments (e.g., Mashup creation and cloud computing). In the past, various methods have been proposed to facilitate the service selection process based on the original functional descriptions. However, the mined features from the descriptions are usually too sparse for training a well-performed model. In addition, most methods neglect to differentiate the weights of various features, while words included in descriptions usually exhibit different intentions (e.g., functional or non-functional). To address these challenges, in this paper we propose a text expansion and deep model-based approach for service recommendation. Specifically, we first expand the description of services at sentence level based on a novel probabilistic topic model that learns topics of words, sentences and descriptions in a stratified fashion. The expansion process can bridge the vocabulary gap between services and user queries with the collective semantic similarity of sentences and descriptions. Then, we propose a Long Short-Term Memory-based model to recommend services with two attention mechanisms - a functional attention mechanism that takes tags as functional prior to mine the function-related features of services and Mashups, and a contextual attention mechanism that considers Mashup requirements as application scenario to help select the most appropriate services. We evaluate the proposed approach on a real-world dataset and the results show it has an improvement of 34 percent in F-measure over the basic LSTM model.

Due to the rapid growth in both the number and diversity of Web services on the web, it becomes increasingly difficult for us to find the desired and appropriate Web services nowadays. Clustering Web services according to their functionalities becomes an efficient way to facilitate the Web services discovery as well as the services management. Existing methods for Web services clustering mostly focus on utilizing directly key features from WSDL documents, e.g., input/output parameters and keywords from description text. Probabilistic topic model Latent Dirichlet Allocation (LDA) is also adopted, which extracts latent topic features of WSDL documents to represent Web services, to improve the accuracy of Web services clustering. However, the power of the basic LDA model for clustering is limited to some extent. Some auxiliary features can be exploited to enhance the ability of LDA. Since the word vectors obtained by Word2vec is with higher quality than those obtained by LDA model, we propose, in this paper, an augmented LDA model (named WE-LDA) which leverages the high-quality word vectors to improve the performance of Web services clustering. In WE-LDA, the word vectors obtained by Word2vec are clustered into word clusters by K-means++ algorithm and these word clusters are incorporated to semi-supervise the LDA training process, which can elicit better distributed representations of Web services. A comprehensive experiment is conducted to validate the performance of the proposed method based on a ground truth dataset crawled from ProgrammableWeb. Compared with the state-of-the-art, our approach has an average improvement of 5.3% of the clustering accuracy with various metrics.

The rapid growth in the number and diversity of Mashup services, coupled with the myriad of functionally similar Mashup services, makes it difficult to find suitable Mashup services to develop Mashup-based software applications due to an unprecedentedly large number of choices of Mashup services. Even if the existing latent factor based methods show significant improvements in Mashup service clustering and discovery, it is still challenging to find Mashup services with high accuracy due to overlooking of relationships among Mashup services. The relationships among Mashup services actually can be exploited in mining latent functional factors to improve the accuracy of clustering and discovery. In this paper, we propose a Mashup service clustering method based on an integration of service content and network via exploiting a two-level topic model. This method, firstly designs a two-level topic model to mine latent topics for representing functional features of Mashup services. Secondly, it uses two different random walk processes to derive and incorporate the topic distribution of Mashup services at service network level into the topic distribution of Mashup services at the service content level. Thirdly, K-means and Agnes algorithm are used to perform Mashup service clustering based on latent topics' similarity. Finally, we conduct a comprehensive evaluation to measure performance of our method. Compared with other existing clustering approaches, experimental results show that our approach achieves a significant improvement in terms of precision, recall, purity and entropy.

Mashups are prevalent Service-Oriented Architecture (SOA) based applications consisting of multiple Web Application Programming Interfaces (APIs) and content. Tags have been extensively used to organize and index mashup services. However, people favor manual tags creation in the past. This approach demands user intervention, which is extremely time-consuming and probes to errors. In this paper we propose a novel Mashup-API-Tag model for automatic mashup tag recommendation. The model simultaneously incorporates the composition relationships between mashups and APIs as well as the annotation relationships between APIs and tags to discover the latent topics. Then the semantic similarity between Web APIs and mashups can be acquired. Subsequently, tags of chosen APIs are recommended to a mashup where the mashup and the APIs are most similar. In addition, we develop a tag filtering algorithm to select the most relevant tags for recommendation. The experimental results on a real world dataset prove that our approach outperforms other methods, including frequency-based methods and the methods that only consider the composition relationships and the annotation relationships separately.