Title: Tensor Eigenvectors and Stochastic Processes
Speakers: A. Benson, Cornell University and D. Gleich, Purdue University
Abstract: The current generation of tensor analysis and computations has been a significant success story for studying datasets now routinely collected in diverse scientific disciplines such as signal processing, biology, and social networks. In this tutorial, the presenters will cover recent innovations and relationships between tensor eigenvectors and stochastic processes that present a challenging new set of computational motivations, generalizations, and trade-offs. Much of the ALA community is familiar with the close relationships between Markov chains—a simple type of stochastic process—and matrix computations. For instance, stationary distributions of Markov chains can be formulated as the solution to a matrix eigenvector problem. By the end of this mini-tutorial, we will have “gone up” a dimension and surveyed relationships between tensors, higher-order Markov chains, and various types of tensor eigenvectors, as well as their applications in data analysis. To this end, the tutorial will involve new types of stochastic processes including vertex-reinforced random walks and spacey random walks. This line of research area has raised several possibilities for future research, and this tutorial will place a specific emphasis on providing well-defined open problems.