### Abstract

A two class pattern recognition algorithm is introduced for learning the optimal separating function within a given finite dimensional linear space of functions. It is proved that the algorithm converges with probability one to the separating function minimizing the probability of misclassification over the given class. This stochastic gradient process is based on asymptotically unbiased estimates of the gradient vector of error probability. This method is proposed. to improve classification rules obtained by other methods.

Original language | English |
---|---|

Title of host publication | Probl Control Inf Theory |

Pages | 371-382 |

Number of pages | 12 |

Volume | 5 |

Edition | 4 |

Publication status | Published - 1976 |

### Fingerprint

### ASJC Scopus subject areas

- Engineering(all)

### Cite this

*Probl Control Inf Theory*(4 ed., Vol. 5, pp. 371-382)

**ON THE MINIMIZATION OF CLASSIFICATION ERROR PROBABILITY IN STATISTICAL PATTERN RECOGNITION.** / Fritz, J.; Györfi, L.

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

*Probl Control Inf Theory.*4 edn, vol. 5, pp. 371-382.

}

TY - CHAP

T1 - ON THE MINIMIZATION OF CLASSIFICATION ERROR PROBABILITY IN STATISTICAL PATTERN RECOGNITION.

AU - Fritz, J.

AU - Györfi, L.

PY - 1976

Y1 - 1976

N2 - A two class pattern recognition algorithm is introduced for learning the optimal separating function within a given finite dimensional linear space of functions. It is proved that the algorithm converges with probability one to the separating function minimizing the probability of misclassification over the given class. This stochastic gradient process is based on asymptotically unbiased estimates of the gradient vector of error probability. This method is proposed. to improve classification rules obtained by other methods.

AB - A two class pattern recognition algorithm is introduced for learning the optimal separating function within a given finite dimensional linear space of functions. It is proved that the algorithm converges with probability one to the separating function minimizing the probability of misclassification over the given class. This stochastic gradient process is based on asymptotically unbiased estimates of the gradient vector of error probability. This method is proposed. to improve classification rules obtained by other methods.

UR - http://www.scopus.com/inward/record.url?scp=0017268116&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0017268116&partnerID=8YFLogxK

M3 - Chapter

VL - 5

SP - 371

EP - 382

BT - Probl Control Inf Theory

ER -