Statistics, Computer Science, Finance and other Shenanigans
From Neural Networks going Deep to Regression to the mean it is probably the first time in history that statistics use has been so prolific in real life. Mobile phone companies use Neural Networks to model the 3-dimensional structure of people’s faces for authentication. Supermarket conglomerates use random effect models to predict customer demand for toothbrushes. It is mesmerizing to imagine how the accelerating trend will reshape everyday life as we know it.
In this talk I would try to give the perspective of a computer literate statistician, nowadays called data scientist, surfing this wind of change and confronting real life problems ranging from finance to shipping.
Dimitris Tasoulis received his Ph.D. from the Department of Mathematics of the University of Patras Greece, where he also completed his under and post graduate studies.
Since then he has served as a Lecturer at the Mathematics Department of Imperial College in London and held a Senior Vice President position at Winton Capital in the finance industry. His most recent post is Head of Algorithmic Trading in the Signal Ocean company that exercises in the maritime and commodity trading fields.
He has published articles in very diverse fields ranging from Data Analysis, Machine Learning, Optimization and Cryptography as well as Bioinfomatics. His work has been cited numerous times with a google scholar calculated h-index of 24.
Symmetrically Adapted Data Summaries
Experimental data are always obtained through measurements which in turn are scalar-valued functions defined for certain labels as points in a functional domain. When the domain has a (set, geometric, algebraic, relational) structure amenable to automorphism groups it is often possible to define measurement summaries that reduce according to the well-known building-blocks properties of those finite groups when probed with the group symmetries. These are the symmetrically adapted summaries of interest. When coupled with sampling these measurements are endowed with classical decompositions of sum of squares and cross-products – the consequence of the canonical projections theorem for finite groups. This talk has the purpose of emphasizing the methodological usefulness and interpretations of these data summaries in a selected number of applications and experimental settings. (Slides can be accessed from https://s3-us-west-2.