PAC learnability under non-atomic measures: a problem by Vidyasagar

05/27/2011
by   Vladimir Pestov, et al.
0

In response to a 1997 problem of M. Vidyasagar, we state a criterion for PAC learnability of a concept class C under the family of all non-atomic (diffuse) measures on the domain Ω. The uniform Glivenko--Cantelli property with respect to non-atomic measures is no longer a necessary condition, and consistent learnability cannot in general be expected. Our criterion is stated in terms of a combinatorial parameter ( C mod ω_1) which we call the VC dimension of C modulo countable sets. The new parameter is obtained by "thickening up" single points in the definition of VC dimension to uncountable "clusters". Equivalently, ( Cω_1)≤ d if and only if every countable subclass of C has VC dimension ≤ d outside a countable subset of Ω. The new parameter can be also expressed as the classical VC dimension of C calculated on a suitable subset of a compactification of Ω. We do not make any measurability assumptions on C, assuming instead the validity of Martin's Axiom (MA). Similar results are obtained for function learning in terms of fat-shattering dimension modulo countable sets, but, just like in the classical distribution-free case, the finiteness of this parameter is sufficient but not necessary for PAC learnability under non-atomic measures.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
08/19/2023

Computing the Vapnik Chervonenkis Dimension for Non-Discrete Settings

In 1984, Valiant [ 7 ] introduced the Probably Approximately Correct (PA...
research
03/03/2022

A Characterization of Multiclass Learnability

A seminal result in learning theory characterizes the PAC learnability o...
research
02/12/2018

Dimension-free PAC-Bayesian bounds for the estimation of the mean of a random vector

In this paper, we present a new estimator of the mean of a random vector...
research
07/07/2023

Optimal Learners for Realizable Regression: PAC Learning and Online Learning

In this work, we aim to characterize the statistical complexity of reali...
research
08/11/2020

Conditions for the existence of a generalization of Rényi divergence

We give necessary and sufficient conditions for the existence of a gener...
research
01/14/2021

New bounds for k-means and information k-means

In this paper, we derive a new dimension-free non-asymptotic upper bound...
research
08/22/2023

Graph-like Scheduling Problems and Property B

Breuer and Klivans defined a diverse class of scheduling problems in ter...

Please sign up or login with your details

Forgot password? Click here to reset