On-line learning with restricted training sets: Exact solution as benchmark for general theories

M S Kearns; S A Solla; D A Cohn

On-line learning with restricted training sets: Exact solution as benchmark for general theories

M S Kearns, S A Solla, D A Cohn (Editor)

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

2 Citations (Scopus)

Abstract

We solve the dynamics of on-line Hebbian learning in perceptrons exactly, for the regime where the size of the training set scales linearly with the number of inputs. We consider both noiseless and noisy teachers. Our calculation cannot be extended to non-Hebbian rules, but the solution provides a nice benchmark to test more general and advanced theories for solving the dynamics of learning with restricted training sets.

Original language	English
Title of host publication	Advances in Neural Information Processing Systems
Place of Publication	CAMBRIDGE
Publisher	M I T PRESS
Pages	316 - 322
Number of pages	7
ISBN (Print)	0-262-11245-0
Publication status	Published - 1999
Event	12th Annual Conference on Neural Information Processing Systems (NIPS) - DENVER, COLORADO Duration: 1 Jan 1999 → …

Publication series

Name	ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS

Conference

Conference	12th Annual Conference on Neural Information Processing Systems (NIPS)
City	DENVER, COLORADO
Period	1/01/1999 → …

Cite this

@inbook{2d8e0eb880494d1c9da8793dcbc6cdb8,

title = "On-line learning with restricted training sets: Exact solution as benchmark for general theories",

abstract = "We solve the dynamics of on-line Hebbian learning in perceptrons exactly, for the regime where the size of the training set scales linearly with the number of inputs. We consider both noiseless and noisy teachers. Our calculation cannot be extended to non-Hebbian rules, but the solution provides a nice benchmark to test more general and advanced theories for solving the dynamics of learning with restricted training sets.",

author = "Kearns, {M S} and Solla, {S A} and Cohn, {D A}",

year = "1999",

language = "English",

isbn = "0-262-11245-0",

series = "ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS",

publisher = "M I T PRESS",

pages = "316 -- 322",

booktitle = "Advances in Neural Information Processing Systems",

note = "12th Annual Conference on Neural Information Processing Systems (NIPS) ; Conference date: 01-01-1999",

}

Kearns, MS , Solla, SA & Cohn, DA (ed.) 1999, On-line learning with restricted training sets: Exact solution as benchmark for general theories. in Advances in Neural Information Processing Systems. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS, M I T PRESS, CAMBRIDGE, pp. 316 - 322, 12th Annual Conference on Neural Information Processing Systems (NIPS), DENVER, COLORADO, 1/01/1999.

On-line learning with restricted training sets: Exact solution as benchmark for general theories. / Kearns, M S ; Solla, S A; Cohn, D A (Editor).
Advances in Neural Information Processing Systems. CAMBRIDGE: M I T PRESS, 1999. p. 316 - 322 (ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS).

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

TY - CHAP

T1 - On-line learning with restricted training sets: Exact solution as benchmark for general theories

AU - Kearns, M S

AU - Solla, S A

A2 - Cohn, D A

PY - 1999

Y1 - 1999

N2 - We solve the dynamics of on-line Hebbian learning in perceptrons exactly, for the regime where the size of the training set scales linearly with the number of inputs. We consider both noiseless and noisy teachers. Our calculation cannot be extended to non-Hebbian rules, but the solution provides a nice benchmark to test more general and advanced theories for solving the dynamics of learning with restricted training sets.

AB - We solve the dynamics of on-line Hebbian learning in perceptrons exactly, for the regime where the size of the training set scales linearly with the number of inputs. We consider both noiseless and noisy teachers. Our calculation cannot be extended to non-Hebbian rules, but the solution provides a nice benchmark to test more general and advanced theories for solving the dynamics of learning with restricted training sets.

M3 - Conference paper

SN - 0-262-11245-0

T3 - ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS

SP - 316

EP - 322

BT - Advances in Neural Information Processing Systems

PB - M I T PRESS

CY - CAMBRIDGE

T2 - 12th Annual Conference on Neural Information Processing Systems (NIPS)

Y2 - 1 January 1999

ER -

On-line learning with restricted training sets: Exact solution as benchmark for general theories

Abstract

Publication series

Conference

Fingerprint

Cite this