A simple randomized algorithm for sequential prediction of ergodic time series

L. Györfi, Gabor Lugosi, G. Morvai

Research output: Contribution to journalArticle

27 Citations (Scopus)

Abstract

We present a simple randomized procedure for the prediction of a binary sequence. The algorithm uses ideas from recent developments of the theory of the prediction of individual sequences. We show that if the sequence is a realization of a stationary and ergodic random process then the average number of mistakes converges, almost surely, to that of the optimum, given by the Bayes predictor. The desirable finite-sample properties of the predictor are illustrated by its performance for Markov processes. In such cases the predictor exhibits near-optimal behavior even without knowing the order of the Markov process. Prediction with side information is also considered.

Original languageEnglish
Pages (from-to)2642-2650
Number of pages9
JournalIEEE Transactions on Information Theory
Volume45
Issue number7
DOIs
Publication statusPublished - 1999

Fingerprint

time series
Time series
Markov processes
Binary sequences
Random processes
performance

Keywords

  • Ergodic processes
  • Markov processes
  • On-line learning
  • Sequential prediction
  • Universal prediction

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Information Systems

Cite this

A simple randomized algorithm for sequential prediction of ergodic time series. / Györfi, L.; Lugosi, Gabor; Morvai, G.

In: IEEE Transactions on Information Theory, Vol. 45, No. 7, 1999, p. 2642-2650.

Research output: Contribution to journalArticle

@article{ec47644231ba42d49fd97fc2f7743379,
title = "A simple randomized algorithm for sequential prediction of ergodic time series",
abstract = "We present a simple randomized procedure for the prediction of a binary sequence. The algorithm uses ideas from recent developments of the theory of the prediction of individual sequences. We show that if the sequence is a realization of a stationary and ergodic random process then the average number of mistakes converges, almost surely, to that of the optimum, given by the Bayes predictor. The desirable finite-sample properties of the predictor are illustrated by its performance for Markov processes. In such cases the predictor exhibits near-optimal behavior even without knowing the order of the Markov process. Prediction with side information is also considered.",
keywords = "Ergodic processes, Markov processes, On-line learning, Sequential prediction, Universal prediction",
author = "L. Gy{\"o}rfi and Gabor Lugosi and G. Morvai",
year = "1999",
doi = "10.1109/18.796420",
language = "English",
volume = "45",
pages = "2642--2650",
journal = "IEEE Transactions on Information Theory",
issn = "0018-9448",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "7",

}

TY - JOUR

T1 - A simple randomized algorithm for sequential prediction of ergodic time series

AU - Györfi, L.

AU - Lugosi, Gabor

AU - Morvai, G.

PY - 1999

Y1 - 1999

N2 - We present a simple randomized procedure for the prediction of a binary sequence. The algorithm uses ideas from recent developments of the theory of the prediction of individual sequences. We show that if the sequence is a realization of a stationary and ergodic random process then the average number of mistakes converges, almost surely, to that of the optimum, given by the Bayes predictor. The desirable finite-sample properties of the predictor are illustrated by its performance for Markov processes. In such cases the predictor exhibits near-optimal behavior even without knowing the order of the Markov process. Prediction with side information is also considered.

AB - We present a simple randomized procedure for the prediction of a binary sequence. The algorithm uses ideas from recent developments of the theory of the prediction of individual sequences. We show that if the sequence is a realization of a stationary and ergodic random process then the average number of mistakes converges, almost surely, to that of the optimum, given by the Bayes predictor. The desirable finite-sample properties of the predictor are illustrated by its performance for Markov processes. In such cases the predictor exhibits near-optimal behavior even without knowing the order of the Markov process. Prediction with side information is also considered.

KW - Ergodic processes

KW - Markov processes

KW - On-line learning

KW - Sequential prediction

KW - Universal prediction

UR - http://www.scopus.com/inward/record.url?scp=0033321311&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033321311&partnerID=8YFLogxK

U2 - 10.1109/18.796420

DO - 10.1109/18.796420

M3 - Article

AN - SCOPUS:0033321311

VL - 45

SP - 2642

EP - 2650

JO - IEEE Transactions on Information Theory

JF - IEEE Transactions on Information Theory

SN - 0018-9448

IS - 7

ER -