The fortnightly BRAG meeting will be held this Thursday 28/07/22 at 1 pm via Zoom/GP-Y801. This week we will have presentations by @Aiden Price and @Matthew Adams.
Zoom link: https://qut.zoom.us/j/98910888234?pwd=STZDUEt0R3d1YzRRd0RLczdpeTdodz09
Password: brag@QUT (if prompted)
Aiden’s Talk
Title: Counting Boats – A Resurgence
Abstract: What started in late 2020 as a fun but minor part of a greater project has now turned into its own (albeit small) project in object classification. In this presentation, I would like to step through a simple object classification pipeline designed to detect boats from low resolution satellite images. Along the way, I will note the specific need for our work, some of the research opportunities, and show some results indicative of what’s to come!
Matthew’s Talk
Title: Simplex truncations of the multivariate normal distribution
Abstract: When fitting models to data, the most common assumption is that the statistical distribution for the model-data fit is Gaussian. However, sometimes this assumption requires modifications to deal with data type; for example, if we know that the data cannot take negative values. A more complicated, but common, problem arises when the data is composed of multiple fractions (so-called, “compositional data”), such that their sum must be less than or equal to one. The statistical properties of the resulting simplex-truncated multivariate normal distribution (ST-MND) are surprisingly not well characterised. In this talk, I will talk about how to perform computations on ST-MNDs using 3 different methods which together require:
- Elliptical slice sampling, using a Markov chain Monte Carlo algorithm possessing 100% acceptance rate;
- Rare event sampling;
- The inclusion-exclusion principle for convex polyhedra;
- Moment generating functions; and
- Trigonometric identities. (Oh no.)
This work was originally motivated by a desire for rapid likelihood computations on a deterministic ODE model of coral-macroalgae competition, but probably has other applications as well. https://arxiv.org/abs/2204.09254
Thanks,
Katie & Jamie
BRAG Co-Chairs