Roman Garnett


Roman Garnett

Associate Professor
Department of Computer Science and Engineering
Washington University in St. Louis

email: garnett at wustl dot edu
phone: (314) 935–4992
address: Washington University in St. Louis
One Brookings Drive, CB 1045
St. Louis, MO 63130

About


I joined the CSE faculty at Wash U in January 2015. Before arriving here, I was a postdoctoral researcher in the Knowledge Discovery and Machine Learning research group at the University of Bonn (2012–2014) and the Auton Lab at Carnegie Mellon University (2010–2012). I completed my PhD as a member of the Machine Learning Research Group at the University of Oxford in 2010, and received an AB and MSc from Washington University in St. Louis in 2004.

You can find my CV here.

Research


Broadly, my research concerns Bayesian experimental design. I am especially interested in applications to scientific discovery, and received an NSF CAREER award to support this research. Two particular problems of focus are:

    Cover of Bayesian Optimization Textbook
  • active search, where we seek to locate novel members of a rare, valuable class. Active search is a model for many settings in scientific discovery, such as drug discovery.
  • Bayesian optimization, where we seek to globally optimize an expensive objective function. Bayesian optimization has seen widespread successful use across a range of application domains across the sciences and engineering, including in machine learning itself for tuning the hyperparameters of complex models. DeepMind recently used Bayesian optimization to learn architectures as part of AlphaGo Zero.

    I wrote a wrote book on Bayesian optimization! It is available free to download.

Publications


You may find a list of my publications here, or may wish to refer to my Google Scholar profile.

Working with me


Thanks for your interest! Here is some advice for those seeking to work with me:

  • Aspiring Ph.D. students: please note that the WUSTL CSE department considers all admissions collectively in a single pool. Students are initially admitted without a graduate advisor assigned. You can find application instructions here.
  • Master's and undergraduate students at Wash U: if you are seeking to complete a master's thesis, a master's project, or an undergraduate independent study with me, I strongly suggest you take my course on Bayesian machine learning. I have successfully completed numerous projects with both undergraduate and graduate students who have taken the course, and these collaborations have led to several publications.

    If you wish to pursue such a project, please identify a project and a potential plan for completing it before approaching me. These projects are meant to be student led, and you will get much more out of the project if you have a personal connection to it.

  • Undergraduate students at any institution: The CSE department has an active REU program! We are currently focusing on big data analytics, and I have been heavily involved since joining the department. Consider applying!

Teaching


CSE 515T – Bayesian Methods in Machine Learning: Fall 2019 Spring 2019 Spring 2018 Spring 2017 Spring 2015
CSE 511A – Introduction to Artificial Intelligence: Spring 2020 Fall 2018 Fall 2017 Fall 2016 Fall 2015