[GrowNet] Gradient Boosting Neural Networks

Time: Wednesday 17-Jun-2020 16:00 (This is a past event.)

Discussion Facilitator:

Artifacts

Motivation / Abstract
Sarkhan will demonstrate how to build a DNN using gradient boosting shallow networks, and how to outperform other SOTA models in computation time and accuracy.
Questions Discussed
- What happens geometrically when augmenting data with a previous network's penultimate layer?
- How do we prevent mutual information? 
- Comparison of Grownet vs popular DNNs
Subjects:
Stream Categories:
 SpotlightAuthor SpeakingMath and Foundations