Noise model on learning sets of strings

Yasubumi Sakakibara; Rani Siromoney

Noise model on learning sets of strings

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In this paper, we introduce a new noise model on learning sets of strings in the framework of PAC learning and consider the effect of the noise on learning. The instance domain is the set Σⁿ of strings over a finite alphabet Σ, and the examples are corrupted by purely random errors affecting only the instances (and not the labels). We consider three types of errors on instances, called EDIT operation errors. EDIT operations consist of `insertion', `deletion', and `change' of a symbol in a string. We call such a noise where the examples are corrupted by random errors of EDIT operations on instances the EDIT noise. First we show general upper bounds on the EDIT noise rate that a learning algorithm of taking the strategy of minimizing disagreements can tolerate and a learning algorithm can tolerate. Next we present an efficient algorithm that can learn a class of decision lists over the attributes `a string w contains a pattern p?' from noisy examples under some restriction on the EDIT noise rate.

Original language	English
Title of host publication	Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory
Publisher	Publ by ACM
Pages	295-302
Number of pages	8
ISBN (Print)	089791497X
Publication status	Published - 1992 Dec 1
Externally published	Yes
Event	Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory - Pittsburgh, PA, USA Duration: 1992 Jul 27 → 1992 Jul 29

Publication series

Name	Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory

Other

Other	Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory
City	Pittsburgh, PA, USA
Period	92/7/27 → 92/7/29

ASJC Scopus subject areas

Engineering(all)

Cite this

Sakakibara, Y & Siromoney, R 1992, Noise model on learning sets of strings. in Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory. Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory, Publ by ACM, pp. 295-302, Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory, Pittsburgh, PA, USA, 92/7/27.

@inproceedings{81d4a333dec042a280440fa476a0800e,

title = "Noise model on learning sets of strings",

abstract = "In this paper, we introduce a new noise model on learning sets of strings in the framework of PAC learning and consider the effect of the noise on learning. The instance domain is the set Σn of strings over a finite alphabet Σ, and the examples are corrupted by purely random errors affecting only the instances (and not the labels). We consider three types of errors on instances, called EDIT operation errors. EDIT operations consist of `insertion', `deletion', and `change' of a symbol in a string. We call such a noise where the examples are corrupted by random errors of EDIT operations on instances the EDIT noise. First we show general upper bounds on the EDIT noise rate that a learning algorithm of taking the strategy of minimizing disagreements can tolerate and a learning algorithm can tolerate. Next we present an efficient algorithm that can learn a class of decision lists over the attributes `a string w contains a pattern p?' from noisy examples under some restriction on the EDIT noise rate.",

author = "Yasubumi Sakakibara and Rani Siromoney",

year = "1992",

month = dec,

day = "1",

language = "English",

isbn = "089791497X",

series = "Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory",

publisher = "Publ by ACM",

pages = "295--302",

booktitle = "Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory",

note = "Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory ; Conference date: 27-07-1992 Through 29-07-1992",

}

TY - GEN

T1 - Noise model on learning sets of strings

AU - Sakakibara, Yasubumi

AU - Siromoney, Rani

PY - 1992/12/1

Y1 - 1992/12/1

N2 - In this paper, we introduce a new noise model on learning sets of strings in the framework of PAC learning and consider the effect of the noise on learning. The instance domain is the set Σn of strings over a finite alphabet Σ, and the examples are corrupted by purely random errors affecting only the instances (and not the labels). We consider three types of errors on instances, called EDIT operation errors. EDIT operations consist of `insertion', `deletion', and `change' of a symbol in a string. We call such a noise where the examples are corrupted by random errors of EDIT operations on instances the EDIT noise. First we show general upper bounds on the EDIT noise rate that a learning algorithm of taking the strategy of minimizing disagreements can tolerate and a learning algorithm can tolerate. Next we present an efficient algorithm that can learn a class of decision lists over the attributes `a string w contains a pattern p?' from noisy examples under some restriction on the EDIT noise rate.

AB - In this paper, we introduce a new noise model on learning sets of strings in the framework of PAC learning and consider the effect of the noise on learning. The instance domain is the set Σn of strings over a finite alphabet Σ, and the examples are corrupted by purely random errors affecting only the instances (and not the labels). We consider three types of errors on instances, called EDIT operation errors. EDIT operations consist of `insertion', `deletion', and `change' of a symbol in a string. We call such a noise where the examples are corrupted by random errors of EDIT operations on instances the EDIT noise. First we show general upper bounds on the EDIT noise rate that a learning algorithm of taking the strategy of minimizing disagreements can tolerate and a learning algorithm can tolerate. Next we present an efficient algorithm that can learn a class of decision lists over the attributes `a string w contains a pattern p?' from noisy examples under some restriction on the EDIT noise rate.

UR - http://www.scopus.com/inward/record.url?scp=0026987081&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026987081&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0026987081

SN - 089791497X

T3 - Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory

SP - 295

EP - 302

BT - Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory

PB - Publ by ACM

T2 - Proceedings of the Fifth Annual ACM Workshop on Computational Learning Theory

Y2 - 27 July 1992 through 29 July 1992

ER -

Noise model on learning sets of strings

Abstract

Publication series

Other

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this