This research-led course provides an introduction to recent developments in Random Matrix Theory and Online Learning that addresses the challenges and opportunities posed by the availability of large amounts of data.

In the first instance, we will review some classic results from multivariate statistical theory, matrix analysis, and probability theory. Then we will present the salient statistical features of big data (e.g., heterogeneity, noise accumulation, spurious correlation, and incidental endogeneity) and show how this impacts on traditional statistical methods and theory.

We follow with an introduction to modern Random Matrix theory and its application in statistics. Applications presented may include topics such as high-dimensional statistical inference, large covariance matrices, large-scale statistical learning through subsampling, sparsification of large matrices, principal component analysis, and dimension reduction.

We conclude with an introduction to the theory of online learning (aka. sequential prediction) to handle the situation of streaming data.

The weekly 2-hour lectures will be accompanied by a 1-hour practical workshop where students will use and learn about the latest computational tools to work with big and streaming data sets. Example data sets may be drawn from areas such finance, web analytics, digital marketing, and satellite imagery data.