Watching online videos has become more and more popular and users tend to watch videos based on their personal tastes and preferences. Providing a customized ranking list to maximize the user's satisfaction has become increasingly important for online video platforms. Existing personalized search methods (PSMs) train their models with user feedback information (e.g. clicks). However, we identified that such feedback signals may indicate attractiveness but not necessarily indicate relevance in video search. Besides, the click data and user historical information are usually too sparse to train a good PSM, which is different from the conventional Web search containing users' rich historical information. To address these concerns, in this paper we propose a multi-task graph neural network architecture for personalized video search (MGNN-PVS) that can jointly model user's click behaviour and the relevance between queries and videos. To relieve the sparsity problem and learn better representation for users, queries and videos, we develop an efficient and novel GNN architecture based on neighborhood sampling and hierarchical aggregation strategy by leveraging their different hops of neighbors in the user-query and query-document click graph. Extensive experiments on a major commercial video search engine show that our model significantly outperforms state-of-the-art PSMs, which illustrates the effectiveness of our proposed framework.

Almost all the existing network embedding methods learn to map the node IDs to their corresponding node embeddings. This design principle, however, hinders the existing methods from being applied in real cases. Node ID is not generalizable and, thus, the existing methods have to pay great effort in cold-start problem. The heterogeneous network usually requires extra work to encode node types, as node type is not able to be identified by node ID. Node ID carries rare information, resulting in the criticism that the existing methods are not robust to noise. To address this issue, we introduce Compositional Network Embedding, a general inductive network representation learning framework that generates node embeddings by combining node features based on the "principle of compositionally". Instead of directly optimizing an embedding lookup based on arbitrary node IDs, we learn a composition function that infers node embeddings by combining the corresponding node attribute embeddings through a graph-based loss. For evaluation, we conduct the experiments on link prediction under three different settings. The results verified the effectiveness and generalization ability of compositional network embeddings, especially on unseen nodes.

This paper studies the problem of social network embedding without relying on network structures that are usually not observed in many cases. We address that the information diffusion process across networks naturally reflects rich proximity relationships between users. Meanwhile, social networks contain multiple communities regularizing communication pathways for information propagation. Based on the above observations, we propose a probabilistic generative model, called COSINE, to learn community-preserving social network embeddings from the recurrent and time-stamped social contagion logs, namely information diffusion cascades. The learned embeddings therefore capture the high-order user proximities in social networks. Leveraging COSINE, we are able to discover underlying social communities and predict temporal dynamics of social contagion. Experimental results on both synthetic and real-world datasets show that our proposed model significantly outperforms the existing approaches.

Neural network techniques are widely used in network embedding, boosting the result of node classification, link prediction, visualization and other tasks in both aspects of efficiency and quality. All the state of art algorithms put effort on the neighborhood information and try to make full use of it. However, it is hard to recognize core periphery structures simply based on neighborhood. In this paper, we first discuss the influence brought by random-walk based sampling strategies to the embedding results. Theoretical and experimental evidences show that random-walk based sampling strategies fail to fully capture structural equivalence. We present a new method, SNS, that performs network embeddings using structural information (namely graphlets) to enhance its quality. SNS effectively utilizes both neighbor information and local-subgraphs similarity to learn node embeddings. This is the first framework that combines these two aspects as far as we know, positively merging two important areas in graph mining and machine learning. Moreover, we investigate what kinds of local-subgraph features matter the most on the node classification task, which enables us to further improve the embedding quality. Experiments show that our algorithm outperforms other unsupervised and semi-supervised neural network embedding algorithms on several real-world datasets.

Recently, online social networks are becoming increasingly popular platforms for social interactions. Understanding how information propagates in such networks is important for personalization and recommendation in social search.

Community detection is a hot topic for researchers in the fields including graph theory, social networks and biological networks. Generally speaking, a community refers to a group of densely linked nodes in the network. Nodes usually have more than one community label, indicating their multiple roles or functions in the network. Unfortunately, existing solutions aiming at overlapping-community-detection are not capable of scaling to large-scale networks with millions of nodes and edges. In this paper, we propose a fast overlapping-communitydetection algorithm - FOX. In the experiment on a network with 3.9 millions nodes and 20 millions edges, the detection finishes in 14 minutes and provides the most qualified results. The second fastest algorithm, however, takes ten times longer to run. As for another network with 22 millions nodes and 127 millions edges, our algorithm is the only one that can provide an overlapping community detection result and it only takes 238 minutes. Our algorithm draws lessons from potential games, a concept in game theory. We measure the closeness of a node to a community by counting the number of triangles formed by the node and two other nodes form the community. Potential games ensure that the algorithm can reach convergence. We also extend the exploitation of triangle to open-triangle, which enlarges the scale of the detected communities.