Near-Optimal Sample Compression for Nearest Neighbors

Lee Ad Gottlieb; Aryeh Kontorovich; Pinhas Nisnevitch

doi:10.1109/TIT.2018.2822267

Near-Optimal Sample Compression for Nearest Neighbors

Lee Ad Gottlieb
, Aryeh Kontorovich
, Pinhas Nisnevitch

Research output: Contribution to journal › Article › peer-review

19 Scopus citations

Abstract

We present the first sample compression algorithm for nearest neighbors with non-trivial performance guarantees. We complement these guarantees by demonstrating almost matching hardness lower bounds, which show that our performance bound is nearly optimal. Our result yields new insight into margin-based nearest neighbor classification in metric spaces and allows us to significantly sharpen and simplify existing bounds. Some encouraging empirical results are also presented.

Original language	English
Pages (from-to)	4120-4128
Number of pages	9
Journal	IEEE Transactions on Information Theory
Volume	64
Issue number	6
DOIs	https://doi.org/10.1109/TIT.2018.2822267
State	Published - 1 Jun 2018

Keywords

Nearest neighbor methods

ASJC Scopus subject areas

Information Systems
Computer Science Applications
Library and Information Sciences

Access to Document

10.1109/TIT.2018.2822267

Cite this

@article{98c6569a29e844929318960d5c5f9d9a,

title = "Near-Optimal Sample Compression for Nearest Neighbors",

abstract = "We present the first sample compression algorithm for nearest neighbors with non-trivial performance guarantees. We complement these guarantees by demonstrating almost matching hardness lower bounds, which show that our performance bound is nearly optimal. Our result yields new insight into margin-based nearest neighbor classification in metric spaces and allows us to significantly sharpen and simplify existing bounds. Some encouraging empirical results are also presented.",

keywords = "Nearest neighbor methods",

author = "Gottlieb, \{Lee Ad\} and Aryeh Kontorovich and Pinhas Nisnevitch",

note = "Publisher Copyright: {\textcopyright} 1963-2012 IEEE.",

year = "2018",

month = jun,

day = "1",

doi = "10.1109/TIT.2018.2822267",

language = "English",

volume = "64",

pages = "4120--4128",

journal = "IEEE Transactions on Information Theory",

issn = "0018-9448",

publisher = "Institute of Electrical and Electronics Engineers",

number = "6",

}

TY - JOUR

T1 - Near-Optimal Sample Compression for Nearest Neighbors

AU - Gottlieb, Lee Ad

AU - Kontorovich, Aryeh

AU - Nisnevitch, Pinhas

PY - 2018/6/1

Y1 - 2018/6/1

N2 - We present the first sample compression algorithm for nearest neighbors with non-trivial performance guarantees. We complement these guarantees by demonstrating almost matching hardness lower bounds, which show that our performance bound is nearly optimal. Our result yields new insight into margin-based nearest neighbor classification in metric spaces and allows us to significantly sharpen and simplify existing bounds. Some encouraging empirical results are also presented.

AB - We present the first sample compression algorithm for nearest neighbors with non-trivial performance guarantees. We complement these guarantees by demonstrating almost matching hardness lower bounds, which show that our performance bound is nearly optimal. Our result yields new insight into margin-based nearest neighbor classification in metric spaces and allows us to significantly sharpen and simplify existing bounds. Some encouraging empirical results are also presented.

KW - Nearest neighbor methods

UR - https://www.scopus.com/pages/publications/85044760632

U2 - 10.1109/TIT.2018.2822267

DO - 10.1109/TIT.2018.2822267

M3 - Article

AN - SCOPUS:85044760632

SN - 0018-9448

VL - 64

SP - 4120

EP - 4128

JO - IEEE Transactions on Information Theory

JF - IEEE Transactions on Information Theory

IS - 6

ER -

Near-Optimal Sample Compression for Nearest Neighbors

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this