Graph Neural Network (GNN) has been attaching great attention along with its successful industry applications, such as social network, recommendation system and so on. Most existing GNN algorithms for link inference tasks mainly concentrate on homogeneous network where single typed nodes and edges are considered. Besides, they are transductive, incapable of handling unseen data, and are difficult to generalize to big graph data. In this paper, we introduce a new idea, i.e. Heterogeneous Multi-view Graph Neural Network (HMGNN), to remedy these problems. A more complex and unstudied heterogeneous network structure where multiple node and edge types co-exist, and each of them also contains specific attributes, is learned in this framework. The proposed HMGNN is end-to-end and two stages are designed: i) The first stage extends the widely-used GraphSAGE model to the studied heterogeneous scenario to generate the vector embedded representations for each type of nodes. ii) The second stage develops a novel and inductive subspace-based strategy for link inference by aggregating multi-typed node and edge feature views. Comprehensive experiments on large-scale spam detection and link prediction applications clearly verify the effectiveness of our model.

Graph-based semi-supervised classification (GSSC) takes labeled and unlabeled samples as vertices in a graph, and edge weights as the similarity between samples. Most GSSC methods handle each labeled sample as equally important in the graph, and they mainly focus on optimizing the graph to improve the performance. In fact, samples are not always evenly distributed. Labeled samples close to the decision boundary of different classes are generally more important than labeled samples far away from the boundary. To account for the different importances, we propose an approach called Weighted Samples based Semi-Supervised Classification (WS3C for short). WS3C firstly executes multiple clusterings on the dataset to explore the structure of samples and summarizes these clustering results. Second, it quantifies the hard-to-cluster index of each labeled sample with respect to other samples based on the summarized results and employs the index to weight that sample. Next, it constructs a graph whose edge weights are equal to the frequency of two samples grouped into the same clusters in multiple clusterings. After that, it performs semi-supervised classification based on the constructed graph and weighted samples. Empirical study on synthesized and real datasets demonstrates that assigning labeled samples with different weights significantly improves the accuracy than equally treating labeled samples. WS3C not only has better performance than other related comparing methods, but also is robust to the input parameters.

Heterogeneous network embedding (HNE) is a challenging task due to the diverse node types and/or diverse relationships between nodes. Existing HNE methods are typically unsupervised. To maximize the profit of utilizing the rare and valuable supervised information in HNEs, we develop a novel Active Heterogeneous Network Embedding (ActiveHNE) framework, which includes two components: Discriminative Heterogeneous Network Embedding (DHNE) and Active Query in Heterogeneous Networks (AQHN). In DHNE, we introduce a novel semi-supervised heterogeneous network embedding method based on graph convolutional neural network. In AQHN, we first introduce three active selection strategies based on uncertainty and representativeness, and then derive a batch selection method that assembles these strategies using a multi-armed bandit mechanism. ActiveHNE aims at improving the performance of HNE by feeding the most valuable supervision obtained by AQHN into DHNE. Experiments on public datasets demonstrate the effectiveness of ActiveHNE and its advantage on reducing the query cost.

Heterogeneous network embedding (HNE) is a challenging task due to the diverse node types and/or diverse relationships between nodes. Existing HNE methods are typically unsupervised. To maximize the profit of utilizing the rare and valuable supervised information in HNEs, we develop a novel Active Heterogeneous Network Embedding (ActiveHNE) framework, which includes two components: Discriminative Heterogeneous Network Embedding (DHNE) and Active Query in Heterogeneous Networks (AQHN). In DHNE, we introduce a novel semi-supervised heterogeneous network embedding method based on graph convolutional neural networks. In AQHN, we first introduce three active selection strategies based on uncertainty and representativeness, and then derive a batch selection method that assembles these strategies using a multi-armed bandit mechanism. ActiveHNE aims at improving the performance of HNE by feeding the most valuable supervision obtained by AQHN into DHNE. Experiments on public datasets demonstrate the effectiveness of ActiveHNE and its advantage on reducing the query cost.

E-commerce taxonomy plays an essential role in online retail business. Existing taxonomy of e-commerce platforms organizes items into an ontology structure. However, the ontology-driven approach is subject to costly manual maintenance and often does not capture user’s search intention, particularly when user searches by her personalized needs rather than a universal definition of the items. Observing that search queries can effectively express user’s intention, we present a novel large-Scale Hierarchical taxOnomy via grAph based query coaLition (SHOAL) to bridge the gap between item taxonomy and user search intention. SHOAL organizes hundreds of millions of items into a hierarchical topic structure. Each topic that consists of a cluster of items denotes a conceptual shopping scenario, and is tagged with easy-to-interpret descriptions extracted from search queries. Furthermore, SHOAL establishes correlation between categories of ontology-driven taxonomy, and offers opportunities for explainable recommendation. The feedback from domain experts shows that SHOAL achieves a precision of 98% in terms of placing items into the right topics, and the result of an online A/B test demonstrates that SHOAL boosts the Click Through Rate (CTR) by 5%. SHOAL has been deployed in Alibaba and supports millions of searches for online shopping per day. PVLDB Reference Format: Zhao Li, Xia Chen, Xuming Pan, Pengcheng Zou, Yuchen Li and Guoxian Yu. SHOAL: Large-scale Hierarchical Taxonomy via Graph-based Query Coalition in E-commerce. PVLDB, 12(12): 1858-1861, 2019. DOI: https://doi.org/10.14778/3352063.3352084

In e-commerce platforms, mining temporal characteristics in user behavior is conducive to recommend the right product for the user at the right time. Recently, recurrent neural networks (RNNs) based methods have achieved profitable performance in exploring temporal features, however, in complex e-commerce scenarios, user preferences changing over time have not been fully exploited. In order to fill the gap, we propose a novel representation for user preferences with the inspiration of a quantum concept, density matrix. It encodes a mixture of item subspaces and represents distribution of user preferences at one time stamp. Further, such a representation and RNNs are combined to form our proposed Density Matrix based Preference Evolution Networks (DMPENs). Experiments on Amazon datasets as well as real-world e-commerce datasets demonstrate the effectiveness of the proposed methods, which achieve rapid convergence and superior performance compared with the state-of-the-art methods in terms of AUC and accuracy.

Current efforts on multi-label learning generally assume that the given labels of training instances are noise-free. However, obtaining noise-free labels is quite difficult and often impractical, and the presence of noisy labels may compromise the performance of multi-label learning. Partial multi-label learning (PML) addresses the scenario in which each instance is annotated with a set of candidate labels, of which only a subset corresponds to the ground-truth. The PML problem is more challenging than partial-label learning, since the latter assumes that only one label is valid and may ignore the correlation among candidate labels. To tackle the PML challenge, we introduce a feature induced PML approach called fPML, which simultaneously estimates noisy labels and trains multi-label classifiers. In particular, fPML simultaneously factorizes the observed instance-label association matrix and the instance-feature matrix into low-rank matrices to achieve coherent low-rank matrices from the label and the feature spaces, and a low-rank label correlation matrix as well. The low-rank approximation of the instance-label association matrix is leveraged to estimate the association confidence. To predict the labels of unlabeled instances, fPML learns a matrix that maps the instances to labels based on the estimated association confidence. An empirical study on public multi-label datasets with injected noisy labels, and on archived proteomic datasets, shows that fPML can more accurately identify noisy labels than related solutions, and consequently can achieve better performance on predicting labels of instances than competitive methods.

Automatically annotating functions of proteins is a key task in bioinformatics.Functional annotations of proteins are collected from multiple sources; thus, noisy annotations are inevitably introduced.However, the current research in protein function prediction almost always focuses on predicting functions for completely unannotated (or incompletely annotated) proteins, and seldom identifies the noisy annotations of proteins.In this paper, we propose a method called identifying noisy functional annotations (NFAs) of proteins using sparse semantic similarity.NFA first utilizes a protein-function association matrix to store the functional annotations of proteins, differentially weighs the annotations using the evidence codes attached with these annotations, and subsequently upward propagates the weights to the expanded annotations via the hierarchical structure among the functional labels.Next, NFA measures the semantic similarity between proteins by the l1-norm regularized sparse representation on the weighted protein-function association matrix.Finally, it identifies the noisy functions of a protein based on the functions annotated to its semantic neighborhood proteins.The experimental results on two model species (A.thaliana and S.cerevisiae) show that the NFA more accurately identifies noisy annotations than other related methods.Additionally, removing the identified noisy annotations improves the accuracy of the current function prediction model.

Current effort on multi-label learning generally assumes that the given labels are noise-free. However, obtaining noise-free labels is quite difficult and often impractical. In this paper, we study how to identify a subset of relevant labels from a set of candidate ones given as annotations to instances, and introduce a matrix factorization based method called MF-INL. It first decomposes the original instance-label association matrix into two low-rank matrices using nonnegative matrix factorization with feature-based and label-based constraints to retain the geometric structure of instances and label correlations. MF-INL then reconstructs the association matrix using the product of the decomposed matrices, and identifies associations with the lowest confidence as noisy associations. An empirical study on real-world multi-label datasets with injected noisy labels shows that MF-INL can identify noisy labels more accurately than other related solutions and is robust to input parameters. We empirically demonstrate that both feature-based and label-based constraints contribute to boosting the performance of MF-INL.

Multi-label active learning addresses the scarce labeled example problem by querying the most valuable unlabeled examples, or example-label pairs, to achieve a better performance with limited query cost. Current multi-label active learning methods require the scrutiny of the whole example in order to obtain its annotation. In contrast, one can find positive evidence with respect to a label by examining specific patterns (i.e., subexample), rather than the whole example, thus making the annotation process more efficient. Based on this observation, we propose a novel two-stage cost effective multi-label active learning framework, called CMAL. In the first stage, a novel example-label pair selection strategy is introduced. Our strategy leverages label correlation and label space sparsity of multi-label examples to select the most uncertain example-label pairs. Specifically, the unknown relevant label of an example can be inferred from the correlated labels that are already assigned to the example, thus reducing the uncertainty of the unknown label. In addition, the larger the number of relevant examples of a particular label, the smaller the uncertainty of the label is. In the second stage, CMAL queries the most plausible positive subexample-label pairs of the selected example-label pairs. Comprehensive experiments on multi-label datasets collected from different domains demonstrate the effectiveness of our proposed approach on cost effective queries. We also show that leveraging label correlation and label sparsity contribute to saving costs.

Graph plays crucially important roles in graph-based semi-supervised learning (SSL). Most SSL methods construct a single graph over all instances to explore the manifold structure of instances, and then enforce the smoothness constraint over such graph. However, instances in the real world are not always evenly distributed. Some instances from different classes but close to decision boundary may be close to each other, and thus they are easy to be misclassified. To mitigate this issue, we propose an approach called semi-supervised classification based on clustering adjusted similarity (SSC-CAS). SSC-CAS firstly takes advantage of clustering on both labeled and unlabeled instances to explore the global structure and discrimination of instances, and then quantifies the similarity between pairwise cluster centers. Second, it adjusts the similarity between pairwise instances by multiplying the similarity between centers of clusters they belong to. In this way, if two instances are from different clusters, the similarity between them is reduced; otherwise, unchanged. After that, SSC-CAS performs graph-based semi-supervised classification on the graph constructed by the adjusted similarity. Empirical study on both synthetic and UCI data-sets demonstrates that SSC-CAS not only has better performance than other related comparing methods, but also is robust to the input parameters.

Annotating remote sensing images is a challenging task for its labor demanding annotation process and requirement of expert knowledge, especially when images can be annotated with multiple semantic concepts (or labels). To automatically annotate these multi-label images, we introduce an approach called Multi-Label Classification based on Low Rank Representation (MLC-LRR). MLC-LRR firstly utilizes low rank representation in the feature space of images to compute the low rank constrained coefficient matrix, then it adapts the coefficient matrix to define a feature-based graph and to capture the global relationships between images. Next, it utilizes low rank representation in the label space of labeled images to construct a semantic graph. Finally, these two graphs are exploited to train a graph-based multi-label classifier. To validate the performance of MLC-LRR against other related graph-based multi-label methods in annotating images, we conduct experiments on a public available multi-label remote sensing images (Land Cover). We perform additional experiments on five real-world multi-label image datasets to further investigate the performance of MLC-LRR. Empirical study demonstrates that MLC-LRR achieves better performance on annotating images than these comparing methods across various evaluation criteria; it also can effectively exploit global structure and label correlations of multi-label images.

Multi-label dimensionality reduction methods often ask for sufficient labeled samples and ignore abundant unlabeled ones. To leverage abundant unlabeled samples and scarce labeled ones, we introduce a method called Semi-supervised Multi-label Linear Discriminant Analysis (SMLDA). SMLDA measures the dependence between pairwise samples in the original space and in the projected subspace to utilize unlabeled samples. After that, it optimizes the target projective matrix by minimizing the distance of within-class samples, whilst maximizing the distance of between-class samples and the dependence term. Extensive empirical study on multi-label datasets shows that SMLDA outperforms other related methods across various evaluation metrics, and the dependence term is an effective alternative to the widely-used smoothness term.