PAC-Bayes Un-Expected Bernstein Inequality.

NeurIPS, (2019): 12180-12191

Cited by: 3|Views10
EI

Abstract:

We present a new PAC-Bayesian generalization bound. Standard bounds contain a L n ⋅ KL n complexity term which dominates unless L n , the empirical error of the learning algorithm's randomized predictions, vanishes. We manage to replace L n by a term which vanishes in many more situations, essentially whenever the employed learning algori...More

Code:

Data:

ZH
Full Text
Bibtex
Your rating :
0

 

Tags
Comments