Google Vizier: A Service for Black-Box Optimization
Any sufficiently complex system starts looking like a black box when it becomes easier to experiment with than to understand. Hence, black-box optimization has become increasingly important as systems have become more complex. In this talk I’ll describe Google Vizier, a Google-internal system for performing black-box optimization that does everything from optimizing machine learning hyperparameters to designing delicious chocolate chip cookies.
Daniel Golovin (CS PhD ‘08) leads the Vizier team in Google Research, right here in Pittsburgh. His research interests include black-box and online optimization, active learning, and large-scale machine learning. He likes chocolate chip cookies.