El Programa de Doctorado en Ingeniería Industrial organiza, los próximos 15 y 16 de mayo, un curso de optimización bayesiana. Será impartido en horario de 10hrs. a 13hrs. en el Seminario del Departamento de Ingeniería de Sistemas y Automática por el profesor Víctor Zavala (Universidad de Wisconsin).
Los ejercicios y/o ejemplos se realizarán en Python a través de Google Colab (en la web). Es necesario llevar un ordenador portátil.
A continuación, se detallan los materiales, el horario y los contenidos del curso.
Lecture Materials
All lecture materials (slides and Jupyter notebooks) can be found here.
Lecture Schedule
May 15th
- 10:00-10:45 - Motivation and Applications of Bayesian Optimization
This will provide a high-level perspective on BO and showcase diverse applications in energy, controls, materials, and chemistry.
- 11:00-11:45 - Data-Driven Modeling using Kernel Models
This will in provide a basic perspective on how kernel methods can be used to build models from data and capture strong nonlinearity. Python examples will be provided to illustrate concepts.
- 12:00-13:00 - Data-Driven Modeling using Gaussian Process Models
This will in provide a basic perspective on how GP models are a class of kernel methods that facilitate data-driven modeling and quantification of uncertainty. Python examples will be provided to illustrate concepts.
May 16th
- 10:00-10:45 - Basic Bayesian Optimization
This will provide basic concepts of BO (exploration vs. exploitation and acquisition functions based on GP models).
- 11:00 - 12:00 - Tutorial Examples
This will provide tutorial examples on how to implement BO using standard Python tools.
- 12:15 - 13:00 - Advanced Bayesian Optimization
This will discussed advanced variants of BO that combine GP and physical knowledge, to use parametric models (e.g., neural networks), and to conduct batch sampling.