Tight Bounds for the Expected Risk of Linear Classifiers and PAC-Bayes Finite-Sample Guarantees
Tight Bounds for the Expected Risk of Linear Classifiers and PAC-Bayes Finite-Sample Guarantees
Jean Honorio,T. Jaakkola
2014 · DBLP: conf/aistats/HonorioJ14
International Conference on Artificial Intelligence and Statistics · 66 Citations
TLDR
This work analyzes the expected risk of linear classiers for a xed weight vector in the \minimax" setting and derives a novel tight exponentialtail bound for sub-Gaussian random variables.