Abstract There has been an increasing interest in developing embedding methods for heterogeneous graph-structured data. The state-of-the-art approaches often adopt a bi-level aggregation scheme, where the first level aggregates information of neighbors belonging to the same type or group, and the second level employs the averaging or attention mechanism to aggregate the outputs of the first level. We find that bi-level aggregation may suffer from a down-weighting issue and overlook individual node information, especially when there is an imbalance in the number of different typed relations. We develop a new simple yet effective single-level aggregation scheme with infomax encoding, named HIME, for unsupervised heterogeneous graph embedding. Our single-level aggregation scheme performs relation-specific transformation to obtain homogeneous embeddings before aggregating information from multiple typed neighbors. Thus, it emphasizes each neighbor’s equal contribution and does not suffer from the down-weighting issue. Extensive experiments demonstrate that HIME consistently outperforms the state-of-the-art approaches in link prediction, node classification, and node clustering tasks.

Modern recommender systems are required to adapt to the change in user preferences and item popularity. Such a problem is known as the temporal dynamics problem, and it is one of the main challenges in recommender system modeling. Different from the popular recurrent modeling approach, we propose a new solution named LeapRec to the temporal dynamic problem by using trajectory-based meta-learning to model time dependencies. LeapRec characterizes temporal dynamics by two complement components named global time leap (GTL) and ordered time leap (OTL). By design, GTL learns long-term patterns by finding the shortest learning path across unordered temporal data. Cooperatively, OTL learns short-term patterns by considering the sequential nature of the temporal data. Our experimental results show that LeapRec consistently outperforms the state-of-the-art methods on several datasets and recommendation metrics. Furthermore, we provide an empirical study of the interaction between GTL and OTL, showing the effects of long- and short-term modeling.

The focal loss has demonstrated its effectiveness in many real-world applications such as object detection and image classification, but its theoretical understanding has been limited so far. In this paper, we first prove that the focal loss is classification-calibrated, i.e., its minimizer surely yields the Bayes-optimal classifier and thus the use of the focal loss in classification can be theoretically justified. However, we also prove a negative fact that the focal loss is not strictly proper, i.e., the confidence score of the classifier obtained by focal loss minimization does not match the true class-posterior probability. This may cause the trained classifier to give an unreliable confidence score, which can be harmful in critical applications. To mitigate this problem, we prove that there exists a particular closed-form transformation that can recover the true class-posterior probability from the outputs of the focal risk minimizer. Our experiments show that our proposed transformation successfully improves the quality of class-posterior probability estimation and improves the calibration of the trained classifier, while preserving the same prediction accuracy.

Graph neural networks (GNNs) have achieved superior performance and gained significant interest in various domains. However, most of the existing GNNs are considered for homogeneous graphs, whereas real-world systems are usually modeled as heterogeneous graphs or heterogeneous information networks (HINs). Designing a GNN to fully capture the rich semantic information of HINs is significantly challenging owing to the heterogeneity and incompatibility of relations in HINs. To address these issues while utilizing the power of GNNs, we propose a novel unsupervised embedding approach, named Projected Graph Relation-Feature Attention Network (PGRA). PGRA is based on three mechanisms: 1) specific-relation projection that projects the representation vector of each node to a relation-specific space, 2) aggregation with a relation-feature attention network that learns salient neighbors in the aggregation by considering the features of the nodes and compatibility between the connected and target relations, 3) an elegantly designed loss function that captures both the first- and second-order proximities between nodes. The results of extensive experiments on seven real-world datasets illustrate that PGRA outperforms the state-of-the-art methods by a large margin.