2016-Analysis of Big Data

Analysis of Big Data 
Organizer and Chair: Francois Bellavance (HEC Montreal) 

LUKE BORNN, Simon Fraser University
From Pixels to Points: Using Tracking Data to Measure Performance in Professional Sports [PDF]
In this talk I will explore how players perform, both individually and as a team, on a basketball court. By blending advanced spatio-temporal models with geography-inspired mapping tools, we are able to understand player skill far better than either individual tool allows. Using optical tracking data consisting of hundreds of millions of observations, I will demonstrate these ideas by characterizing defensive skill and decision making in NBA players. 
ERSHAD BANIJAMALI, University of Waterloo
Generative Mixture of Networks  [PDF]
In this talk I will introduce a generative model by training deep architecture. The model starts with dividing the input data into K clusters and feeding each of them into a separate network. So, there is a big network which consists of K sub-networks. The goal of training each sub-network is to generate samples which have close distribution to the underlying distribution of its assigned training set. Empirical estimation of Maximum Mean Discrepancy (MMD) is used as a measure of distance between these two distributions. Subsequently, an algorithm is employed whose goal is to further train the networks, jointly with updating the clusters of the training set by a non-parametric likelihood estimation. We call this algorithm, Mixture of Networks
Clustering Ultra High-Dimensional Data  [PDF]
Some approaches to clustering ultra high-dimensional data are considered. Each approach is based on a mixture model, and each component thereof is taken as corresponding to a cluster. Some approaches depend on the assumption of an underlying low-dimensional (latent) space while others also draw on shrinkage. Real data are used for illustration.