Michael Trosset

Michael Trosset

Professor, Statistics

Education

  • Ph.D., University of California, Los Angeles, 1989

Research interests

Computational Statistics, especially problems that involve numerical optimization, e.g., the development of tractable formulations of and efficient numerical algorithms for multidimensional scaling and other methods for embedding dissimilarity data.

Statistical Learning, i.e., multivariate data-analytic techniques for nonlinear dimension reduction (manifold learning), classification, and clustering. Current interests include the application of distance geometry to the problem of inferring 3-dimensional molecular structure from distance restraints, and various high-dimensional classification problems in bioinformatics.

Design & Analysis of Computer Experiments, specifically for the purpose of optimizing computationally expensive computer simulations. Current interests include the application of statistical decision theory to computer-assisted robust design.

Stochastic Optimization and Response Surface Methodology, especially for tuning the inputs of highly nonlinear stochastic simulations and estimating the parameters of analytically intractable stochastic processes. Current interests include developing quasi-Newton methods for optimization in the presence of random noise.

Representative publications

Approximate Information Tests on Statistical Submanifolds.(2019)
Michael W. Trosset, Carey E. Priebe

On the Power of Likelihood Ratio Tests in Dimension-Restricted Submodels.(2016)
Michael W. Trosset, Mingyue Gao, Carey E. Priebe

Fast Embedding for JOFC Using the Raw Stress Criterion.(2015)
Vince Lyzinski, Youngser Park, Carey E. Priebe, Michael W. Trosset
Journal of Computational and Graphical Statistics, 26 (4),

Parallel deterministic and stochastic global minimization of functions with very many minima.(2013)
David R. Easterling, Layne T. Watson, Michael L. Madigan, Brent S. Castle & Michael W. Trosset
Computational Optimization and Applications, 57 (2), 469–492

Supplementary Material.(2014)
David Robert Easterling, Layne T. Watson, Michael Madigan, Brent S. Castle, Michael W. Trosset

Fortran 95 implementation of QNSTOP for global and stochastic optimization.(2014)
Brandon Amos, David Robert Easterling, L.T. Watson, B.S. Castle, Michael W. Trosset, William I. Thacker

Adjusting process count on demand for petascale global optimization.(2012)
Masha Sosonkina, Layne T. Watson, Nicholas R. Radcliff, Rafael T.Haftka, Michael W.Trosset
Parallel Computing, 39 (1), 21-35

Direct search and stochastic optimization applied to two nonconvex nonsmooth problems.(2012)
David Robert Easterling, Layne T. Watson, Michael Madigan, Brent S. Castle, Michael W. Trosset
Proceedings of the 2012 Symposium on High Performance Computing,

Interference competition in desert subterranean termites.(2011)
S. C. Jones, Michael W. Trosset
Entomologia Experimentalis et Applicata, 61 (1), 83-90

Euclidean and circum-Euclidean distance matrices: Characterizations and linear preserves.(2010)
Li Chi-Kwong, Milligan Thomas, Trosset Michael W
The electronic journal of linear algebra ELA, 20 (1),

Dissertation Committee Service

Author Dissertation Title Committee
Blaha, Leslie A Dynamic Hebbian-style Model of Configural Learning (December 2010) Townsend, J. (Co-Chair), Busey, T. (Co-Chair), Gold, J,. Trosset, M.