The Annual Data Assimilation Grand Prix

The Annual Data Assimilation Grand Prix#

Welcome to the event of the year - the Annual Data Assimilation Grand Prix! It has it all - statistics! data! excitement! In this chapter, we pay tribute to the contributors and victors of years past.

So how does the Grand Prix work?

Every year, our students get a mysterious timeseries of 2D data. Their task is then to:

  1. Design a model that replicates the dynamics in the time series as well as possible

  2. Create a filter that corrects the prediction of their model

Then, each student submits their model/filter pairing in a race against their colleagues. Participants get a short time series to train their models, and are then evaluated on the training period and an as-yet-unseen validation period. The winners with the three lowest root mean square errors (compared to the synthetic reference) earn eternal glory, a place for their names immortilized in the MORE Hall of Fame, and chocolate.