Stats, Optimization, and Machine Learning Seminar - Rose Yu

Learning from Large-Scale Spatiotemporal Data

In many real-world applications, such as internet of things (IoT), transportation and physics, machine learning is applied to large-scale spatiotemporal data. Such data is often nonlinear, high-dimensional, and demonstrates complex spatial and temporal correlations. In this talk, I will demonstrate how to efficiently learn from such data.  In particular, I will present some recent results on 1) Low-Rank Tensor Regression for spatiotemporal causal inference and 2) Diffusion Convolutional RNNs for spatiotemporal forecasting, applied to real-world traffic and climate data. I will also discuss opportunities and challenges of learning from large-scale spatiotemporal data.