Google Vizier: A service for black-box optimization

Black-box optimization is the task of optimizing an objective function f : X \rightarrow \mathbb{R} with a limited budget for evaluations. The adjective “black-box” means that while we can evaluate f(x) for any x \in X, we have no access to any other information about f, such as gradients or the Hessian.

the morning paper

Google Vizier: a service for black-box optimization Golovin et al., KDD’17

We finished up last week by looking at the role of an internal (or external) experimentation platform. In today’s paper Google remind us that such experimentation is just one form of optimisation. Google Vizier is an internal Google service for optimising pretty much anything. And if you use for example Google’s public HyperTune system (part of the Google Cloud Machine Learning Engine service), then you’re using Vizier under the covers.

The very opening sentences of the abstract are worth dwelling on for a moment here:

Any sufficiently complex system acts as a black box when it becomes easier to experiment with than to understand. Hence, black-box optimization has become increasingly important as systems become more complex.

This is recognising something important happening all around us: we’ve made systems of sufficient emergent complexity that we don’t truly understand them…

View original post 1,239 more words

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.