Carmen Isabel Martinez Harris
Recovering Parameters of Truncated Multivariate Gaussian Distributions via Numerical Optimisation
This project aims to develop and test numerical optimisation methods to recover the mean (μ) and covariance (Σ) of a Gaussian distribution from its truncated counterpart. The focus will be on using gradient descent techniques to minimise the difference between known truncated parameters (μ_T, Σ_T) and those predicted from trial values, specifically for rectangular bounds in one and two dimensions.
The first objective is to implement and validate the optimisation approach for accurately recovering (μ, Σ) from truncated parameters. The second is to compare its computational efficiency and accuracy against the ODE-based method proposed by Wang and Li (2024). The third is to assess performance in both univariate and bivariate cases with rectangular truncation.
The outcomes will provide a direct comparison of computational efficiency and accuracy between optimisation and ODE-based approaches. This will help identify the conditions under which optimisation methods offer a more practical and precise solution for parameter recovery in truncated multivariate distributions.
Carmen Isabel Martinez Harris
Queensland University of Technology
Carmen Martinez Harris is a researcher from Queensland University of Technology (QUT) with a background in applying quantitative methods to interdisciplinary challenges. She has completed five research projects across diverse fields, including nanotechnology, lithium-ion batteries, space policy and governance, and Indigenous astronomy.
In recognition of her work, Carmen was selected as a student researcher for CSIRO and received the “Best Presentation Award” at a QUT research showcase. Alongside her research, she was active in the STEM community as a mentor for high school students through the ASPIRE program.
Carmen will begin her Master’s degree next year, concentrating on her primary research interest in astrophysics. She is particularly focused on integrating advanced mathematical and computational modeling to improve the methods used to derive and estimate galaxy properties.
