Bayesian Optimization: Basics & Challenges

We're happy to host our first event of 2020! The talk will begin promptly at 6 and last until about 7.

Please bring photo ID and be able to present your meetup confirmation.

Abstract:
This talk will give a brief introduction to Bayesian Optimization (BO). BO is a global optimization method that explicitly builds a probabilistic surrogate model of the objective function in which the domain knowledge of the search objective can be clearly specified and searches for the optimum of the objective function by iteratively suggesting the next location for evaluation. BO has been widely used in various optimization tasks with expensive objective functions such as hyper-parameter optimization, drug discovery, synthetic gene design. Despite being a highly efficient global optimization method, it has lots of open challenges. This talk will mention a few of the challenges and present the related works from the research community.

Speaker Bio
Zhenwen Dai is a senior research scientist at Spotify. He is a member of the tech research group, working on Bayesian optimization, automated machine learning, probabilistic modeling and variational inference. Previously, he worked as a machine learning scientist at Amazon, Cambridge, UK. Before that, he worked as a postdoctoral researcher at the university of Sheffield with Neil Lawrence, after completing his PhD degree with Jorg Lucke on machine learning at the Goethe University Frankfurt.