Abstract:
Bayesian optimization is a popular strategy for sequentially testing parameters to find optimal behavior for systems, models, or arbitrary functions. In the standard case, Bayesian optimization (also called Sequential Metamodel/Kriging/Model-Based Optimization) assumes no gradient information and a noisy objective. Its goal is not to actually find the optimum, but rather to reach the highest performance possible with a limited observation budget; such methods fall into the class of Sample-Efficient methods for sequential decision making. In this talk, we will introduce Bayesian optimization, some of its extensions and adaptations to expand its applicability, and adjacent methods such as active search. We will review some of the applications in which Bayesian optimization thrives, as well as circumstances where other methods may be preferable. We will survey some of the popular Python tools for conducting Bayesian optimization in Python. This talk will not assume any knowledge of Bayesian optimization or Gaussian processes, and will be focused more on the ideas/outcomes/usage of these tools rather than the theory underlying them (though questions about the theory are welcome).
Bio:
Michael McCourt was a Chicago resident for many years, first doing undergrad at the Illinois Institute of Technology and later some time as a postdoc at Argonne National Laboratory. In between, he finished his PhD in Applied Mathematics at Cornell, and he has been focused on research related to positive definite kernels, Gaussian processes and associated theory and applications. In 2015, he moved to the startup life, joining SF-based startup SigOpt as a research engineer to work on the black-box optimization algorithms underlying their SaaS platform. SigOpt was acquired by Intel in 2020, and Mike still works on SigOpt from within Intel as a senior principal engineer. He is thrilled to be able to speak to the data science community in Chicago and looks forward to meeting everyone!
Web: mikemccourt.github.io
www.pydata.org
PyData is an educational program of NumFOCUS, a 501(c)3 non-profit organization in the United States. PyData provides a forum for the international community of users and developers of data analysis tools to share ideas and learn from each other. The global PyData network promotes discussion of best practices, new approaches, and emerging technologies for data management, processing, analytics, and visualization. PyData communities approach data science using many languages, including (but not limited to) Python, Julia, and R.
PyData conferences aim to be accessible and community-driven, with novice to advanced level presentations. PyData tutorials and talks bring attendees the latest project features along with cutting-edge use cases.
00:00 Welcome!
00:10 Help us add time stamps or captions to this video! See the description for details.
Want to help add timestamps to our YouTube videos to help with discoverability? Find out more here: https://github.com/numfocus/YouTubeVideoTimestamps