How to escape saddlepoints efficiently
Michael Jordan discusses recent results in gradient-based optimization for large-scale data analysis.
![Gradient](https://www.oreilly.com/radar/wp-content/uploads/sites/3/2019/06/pexels-photo-434645_crop-13d13aa4c5a36e029148df4cad227bf4-1.jpg)
This is a keynote highlight from the Artificial Intelligence Conference in San Francisco 2017. Watch the full version of this keynote on Safari.
You can also see other highlights from the event.