Probabilistic Deep Network

This post explains a trick which allows us to convert neural network outputs into probabilities, with no cost to performance, and minimal computational overhead.

In Quantum Mechanics, Heisenberg’s Uncertainty Principle states that there is a fundamental limit to how well one can measure a particle’s position and momentum. In the context of machine learning systems, a similar principle has emerged, but relatinginterpretability and performance.

See more on http://www.computervisionblog.com/2016/06/making-deep-networks-probabilistic-via.html

Leave a Reply

Your e-mail address will not be published. Required fields are marked *

*

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