Wasserstein PAC-Bayes Learning: A Bridge Between Generalisation and Optimisation
Wasserstein PAC-Bayes Learning: A Bridge Between Generalisation and Optimisation
Maxime Haddouche,Benjamin Guedj
2023 · DOI: 10.48550/arXiv.2304.07048
arXiv.org · 5 Citations
TLDR
This work expands the Wasserstein PAC-Bayes framework, and proves, before any training, that the output of an algorithm from Lambert et al. (2022) has a strong asymptotic generalisation ability.