Events

An Overview of Variable Annuities: From Market Practice to Mathematical Fun

Over the past decade, the insurance industry in North America has gradually laid out new standards on the stochastic modeling, risk management and financial reporting of equity-linking insurance products. However, it has been widely known that the current market practice of risk assessment based on Monte Carlo simulations can be very time-consuming and costly. Many emerging problems naturally call for new solutions from both practitioners and academics.

Markov bases -- how to use them and how to compute them

To run Fisher's exact test for independence, we have to find all integer tables with prescribed row sums and column sums.  More generally, to run Fisher's exact test for graphical models, we have to find all n-way tables with prescribed marginals.  Often, it is impossible to list all such tables explicitly, but for statistical purposes it suffices to sample from this set, for example by running an MCMC algorithm using a fixed (finite) set of moves to jump from one table to the next.

Grand Challenges in Computer Vision and Imaging

Parsimonious Modeling of Structured Correlation Matrices

There has been a flurry of activity in the last two decades in reparametrizing  Cholesky factors of  correlation matrices

A Career Adventure in Risk Analytics: How to turn Big Data into Big Money

David Jepson, Director of Risk Analysis & Reporting at Manulife Bank, will outline his career evolution from a student in Operations Research & Statistics at the University of Waterloo to Director of Risk Management at Manulife, and everything in between. There will be particular focus on two key experiences:

The Last Line of Defense: Basel II Expected Credit Loss Models (Probability of Default, Exposure at Default, and Loss Given Default) & Risk-Weighted Assets (RWA)

Detection, attribution and extreme climate and weather events

Networks, Random Graphs and Percolation

This event has been CANCELLED

Robust Covariances: Structured Models and Regularization

In this talk, robust estimation of the covariance matrix is considered whenever constrains are placed on the covariance matrix. Such models are particularly important whenever there is low or insufficient sample support (small n, large p).

Monitoring and Improving Surgical Quality