Distribution Learnability and Robustness
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914849103020032 |
|---|---|
| author | Ben-David, Shai Bie, Alex Kamath, Gautam Lechner, Tosca |
| author_facet | Ben-David, Shai Bie, Alex Kamath, Gautam Lechner, Tosca |
| contents | We examine the relationship between learnability and robust (or agnostic) learnability for the problem of distribution learning. We show that, contrary to other learning settings (e.g., PAC learning of function classes), realizable learnability of a class of probability distributions does not imply its agnostic learnability. We go on to examine what type of data corruption can disrupt the learnability of a distribution class and what is such learnability robust against. We show that realizable learnability of a class of distributions implies its robust learnability with respect to only additive corruption, but not against subtractive corruption.
We also explore related implications in the context of compression schemes and differentially private learnability. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_17814 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Distribution Learnability and Robustness Ben-David, Shai Bie, Alex Kamath, Gautam Lechner, Tosca Machine Learning Data Structures and Algorithms Information Theory Statistics Theory We examine the relationship between learnability and robust (or agnostic) learnability for the problem of distribution learning. We show that, contrary to other learning settings (e.g., PAC learning of function classes), realizable learnability of a class of probability distributions does not imply its agnostic learnability. We go on to examine what type of data corruption can disrupt the learnability of a distribution class and what is such learnability robust against. We show that realizable learnability of a class of distributions implies its robust learnability with respect to only additive corruption, but not against subtractive corruption. We also explore related implications in the context of compression schemes and differentially private learnability. |
| title | Distribution Learnability and Robustness |
| topic | Machine Learning Data Structures and Algorithms Information Theory Statistics Theory |
| url | https://arxiv.org/abs/2406.17814 |