Not All Learnable Distribution Classes are Privately Learnable
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866913142304407552 |
|---|---|
| author | Bun, Mark Kamath, Gautam Mouzakis, Argyris Singhal, Vikrant |
| author_facet | Bun, Mark Kamath, Gautam Mouzakis, Argyris Singhal, Vikrant |
| contents | We give an example of a class of distributions that is learnable up to constant error in total variation distance with a finite number of samples, but not learnable under $(\varepsilon, δ)$-differential privacy with the same target error. This weakly refutes a conjecture of Ashtiani. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_00267 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Not All Learnable Distribution Classes are Privately Learnable Bun, Mark Kamath, Gautam Mouzakis, Argyris Singhal, Vikrant Data Structures and Algorithms Cryptography and Security Machine Learning We give an example of a class of distributions that is learnable up to constant error in total variation distance with a finite number of samples, but not learnable under $(\varepsilon, δ)$-differential privacy with the same target error. This weakly refutes a conjecture of Ashtiani. |
| title | Not All Learnable Distribution Classes are Privately Learnable |
| topic | Data Structures and Algorithms Cryptography and Security Machine Learning |
| url | https://arxiv.org/abs/2402.00267 |