Abstract: Tensor factorization methods are popular now. Directly modeling multi-relational data makes tensors attractive. We propose ParaSketch, a massively parallel tensor factorization algorithm for large tensors. Compress the large tensor into multiple small ones, decompose…
Abstract: Multi-source domain adaptation requires reducing domain discrepancy between source domains and target domains and then assessing domain relevance to determine how much knowledge should be transferred. Most previous approaches ignored domain discrepancies and relevance….
Abstract: Most users of big-data analytics don’t know how dimensionality reduction works. Understanding the quality of a low-dimensional embedding helps choose the best dimensionality reduction algorithm and trust the transformed data. Existing research mostly explores…
Abstract: Community search on various graphs finds query-dependent communities. Intimate-core group (community) search over a weighted graph finds a connected k-core of query nodes with the smallest group weight. State-of-the-art methods refine an answer from…
Abstract: In the age of big data, many graph algorithms must operate in external memory and maintain performance regardless of problem size. Triangle listing algorithms must carefully combine edges from multiple partitions to detect cycles…
Abstract: With the rise of online social networks, identifying or classifying real-life relationships between users has become useful for financial fraud detection. People in different relationships usually give each other meaningful gifts on different dates….
Abstract: Network embedding solves network analytics. Methods focus on single-layered homogeneous or heterogeneous networks. Multilayer networks—heterogeneous networks with multiple edge/relation types—can naturally represent many real-world complex systems. Multilayer network embedding struggles to capture and use…
Abstract: Heterogeneous networks are used as a more powerful, realistic, and generic superclass of homogeneous networks (graphs) because real-world objects and interactions are multi-modal and multi-typed. Representation learning (embedding) has been extensively studied for network…
Abstract: Given a user’s purchase/rating trajectories, sequential recommendation recommends the next few items they’re most likely to buy/review. It helps users choose their favorites. This manuscript uses hybrid associations models (HAM) to generate sequential recommendations…
Abstract: Card fraud costs cardholders and issuers a lot. Machine learning-based methods detect fraud from transaction records. However, manually generating features requires domain knowledge and may be behind fraud, so the online detection system must…
Abstract: Due to the time-evolving nature of many real-world networks, learning low-dimensional topological representations in dynamic environments is gaining attention. Dynamic Network Embedding (DNE) aims to efficiently update node embeddings while preserving network topology at…
Abstract: Machine learning libraries need static and dynamic clustering algorithms. The sliding window model or simpler models have dominated dynamic machine learning and data mining algorithm development. Many real-world applications require arbitrary deletions and insertions….
WhatsApp us