Bio
I’m André Gustavo Carlon, a postdoctoral researcher with a Ph.D. in structural engineering. My work focuses on stochastic optimization, particularly the development and analysis of stochastic gradient methods, including variance reduction and preconditioning techniques. I am interested in applications to optimal experimental design, reliability analysis, and machine learning. I serve as an Associate Editor for Statistics and Computing.
Research Interests
- Optimal Experimental Design
- Stochastic Gradient Descent
- Reliability Engineering
- Machine Learning
- Uncertainty Quantification
- Bayesian Inference
Experience
- 2024– Postdoctoral researcher at Mathematics for Uncertainty Quantification, RWTH-Aachen University
- 2020–2024 Postdoctoral fellow at Stochastic Numerics, KAUST
Education
-
2015–2019
Doctorate of Engineering
Universidade Federal de Santa Catarina (UFSC)
Florianópolis, Brazil -
2013–2015
Masters of Engineering
Universidade Federal de Santa Catarina (UFSC)
Florianópolis, Brazil -
2007–2013
Bachelor of Civil Engineering
Universidade Federal de Santa Catarina (UFSC)
Florianópolis, Brazil
Selected Publications
- *, Espath, L. and Tempone, R. (2024). Approximating Hessian matrices using Bayesian inference: a new approach for quasi-Newton methods in stochastic optimization. Optimization Methods and Software, 39(6), 1352–1382. DOI:10.1080/10556788.2024.2339226
- *, Kroetz, H. M., Torii, A. J., Lopez, R. H., and Miguel, L. F. F. (2022). Risk optimization using the Chernoff bound and stochastic gradient descent. Reliability Engineering & System Safety, 223, 108512. DOI:10.1016/j.ress.2022.108512
- *, Dia, B. M., Espath, L., Lopez, R. H., and Tempone, R. (2020). Nesterov-aided stochastic gradient methods using Laplace approximation for Bayesian design optimization. Computer Methods in Applied Mechanics and Engineering, 363, 112909. DOI:10.1016/j.cma.2020.112909
- *, Lopez, R. H., Espath, L., Miguel, L. F. F., and Beck, A. T. (2019). A stochastic gradient approach for the reliability maximization of passively controlled structures. Engineering Structures, 186, 1–12. DOI:10.1016/j.engstruct.2019.01.121
Open-Source Libraries
MICE
A multi‐iteration stochastic estimator (MICE) for stochastic gradient descent.
Install: pip install mice
BayHess
A quasi-Newton pre-conditioner for stochastic gradient descent based on Bayesian inference.
Install: pip install bayhess
Organization
- 2022-2025 — Member of the organization committee of the Statistical Numerics and Statistical Learning workshop at KAUST.