Abstract
In this work we offer a framework for reasoning about a wide class of
existing objectives in machine learning. We develop a formal correspondence
between this work and thermodynamics and discuss its implications.
Users
Please
log in to take part in the discussion (add own reviews or comments).