Scalable Tensor Mining

Tensors, or multi dimensional arrays, are receiving significant attention due to the various types of data that can be modeled by them; examples include call graphs (sender, receiver, time), knowledge bases (subject, verb, object), 3-dimensional web graphs augmented with anchor texts, to name a few. Scalable tensor mining aims to extract important patterns and anomalies from a large amount of tensor data. In this paper, we provide an overview of scalable tensor mining. We first present main algorithms for tensor mining, and their scalable versions. Next, we describe success stories of using tensors for interesting data mining problems including higher order web analysis, knowledge base mining, network traffic analysis, citation analysis, and sensor data analysis. Finally, we discuss interesting future research directions for scalable tensor mining.