Tensors, time series, and computational sustainability
In many real-world applications, such as transportation, climate science and social networks, machine learning is applied on large-scale time series data. Such data is often high-dimensional, multi-resolution and demonstrates complex dependency structures. Tensors, as generalizations of vectors and matrices, provide a natural and scalable framework for handling data with inherent higher-order structures. The recent renaissance of tensor methods has attracted a considerable amount of attention in the machine learning community. In this talk, I will demonstrate how to efficiently learn from time series data with tensor methods, in both offline and online settings. I will also discuss the issues and challenges of learning from large-scale time series in sustainability applications.