Not All Learnable Distribution Classes are Privately Learnable

被引：0
|
作者：
Bun, Mark [1 ]
Kamath, Gautam [2 ,3 ]
Mouzakis, Argyris [2 ]
Singhal, Vikrant [4 ]
机构：
[1] Boston Univ, Boston, MA 02215 USA
[2] Univ Waterloo, Waterloo, ON, Canada
[3] Vector Inst, Toronto, ON, Canada
[4] OpenDP, Boston, MA USA
来源：
INTERNATIONAL CONFERENCE ON ALGORITHMIC LEARNING THEORY, VOL 237 | 2024年 / 237卷
基金：
加拿大自然科学与工程研究理事会;
关键词：
Differential privacy; distribution learning; lower bounds; SAMPLE COMPLEXITY; BOUNDS;
D O I：
暂无
中图分类号：
TP18 [人工智能理论];
学科分类号：
081104 ; 0812 ; 0835 ; 1405 ;
摘要：
We give an example of a class of distributions that is learnable in total variation distance with a finite number of samples, but not learnable under (epsilon,delta)-differential privacy. This refutes a conjecture of Ashtiani.
引用
收藏
页数：12
相关论文
共 50 条
  • [1] Littlestone Classes are Privately Online Learnable
    Golowich, Noah
    Livni, Roi
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 34 (NEURIPS 2021), 2021, 34
  • [2] Enlarging learnable classes
    Jain, Sanjay
    Koetzing, Timo
    Stephan, Frank
    INFORMATION AND COMPUTATION, 2016, 251 : 194 - 207
  • [3] Splitting of Learnable Classes
    Li, Hongyang
    Stephan, Frank
    GRAMMATICAL INFERENCE: THEORETICAL RESULTS AND APPLICATIONS, ICGI 2010, 2010, 6339 : 109 - 121
  • [4] Classes with easily learnable subclasses
    Jain, S
    Menzel, W
    Stephan, F
    ALGORITHMIC LEARNING THEORY, PROCEEDINGS, 2002, 2533 : 218 - 232
  • [5] Maximal machine learnable classes
    Case, J
    Fulk, MA
    JOURNAL OF COMPUTER AND SYSTEM SCIENCES, 1999, 58 (01) : 211 - 214
  • [6] Classes with easily learnable subclasses
    Jain, S
    Menzel, W
    Stephan, F
    INFORMATION AND COMPUTATION, 2004, 190 (01) : 81 - 99
  • [7] Maximal machine learnable classes
    Univ of Delaware, Newark, United States
    J Comput Syst Sci, 1 (211-214):
  • [8] Mixtures of Gaussians are Privately Learnable with a Polynomial Number of Samples
    Afzali, Mohammad
    Ashtiani, Hassan
    Liaw, Christopher
    INTERNATIONAL CONFERENCE ON ALGORITHMIC LEARNING THEORY, VOL 237, 2024, 237
  • [9] Learnable classes of general combinatory grammars
    Moreau, E
    LOGICAL ASPECTS OF COMPUTATIONAL LINGUISTICS, PROCEEDINGS, 2005, 3492 : 189 - 204
  • [10] A characterization of the language classes OF learnable with correction queries
    Tirnauca, Cristina
    Kobayashi, Satoshi
    THEORY AND APPLICATIONS OF MODELS OF COMPUTATION, PROCEEDINGS, 2007, 4484 : 398 - +
12345