### Abstract

Convergence theorems for some learning algorithms generated by a potential function are dealt with. The kernel function of a reproducing kernel Hilbert space (RKHS) as a potential function is used. The problem of how one may use ″teaching″ several times during the algorithm is considered, along with the case of noisy ″teaching″ . Algorithms are discussed which turned out to be particularly efficient in learning in unambiguous models by potential function type learning algorithms.

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

Title of host publication | Probl Control Inf Theory |

Pages | 247-265 |

Number of pages | 19 |

Volume | 1 |

Edition | 3-4 |

Publication status | Published - 1972 |

### Fingerprint

### ASJC Scopus subject areas

- Engineering(all)

### Cite this

*Probl Control Inf Theory*(3-4 ed., Vol. 1, pp. 247-265)

**CONVERGENCE OF POTENTIAL FUNCTION TYPE LEARNING ALGORITHMS.** / Györfi, L.

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

*Probl Control Inf Theory.*3-4 edn, vol. 1, pp. 247-265.

}

TY - CHAP

T1 - CONVERGENCE OF POTENTIAL FUNCTION TYPE LEARNING ALGORITHMS.

AU - Györfi, L.

PY - 1972

Y1 - 1972

N2 - Convergence theorems for some learning algorithms generated by a potential function are dealt with. The kernel function of a reproducing kernel Hilbert space (RKHS) as a potential function is used. The problem of how one may use ″teaching″ several times during the algorithm is considered, along with the case of noisy ″teaching″ . Algorithms are discussed which turned out to be particularly efficient in learning in unambiguous models by potential function type learning algorithms.

AB - Convergence theorems for some learning algorithms generated by a potential function are dealt with. The kernel function of a reproducing kernel Hilbert space (RKHS) as a potential function is used. The problem of how one may use ″teaching″ several times during the algorithm is considered, along with the case of noisy ″teaching″ . Algorithms are discussed which turned out to be particularly efficient in learning in unambiguous models by potential function type learning algorithms.

UR - http://www.scopus.com/inward/record.url?scp=0015457917&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0015457917&partnerID=8YFLogxK

M3 - Chapter

AN - SCOPUS:0015457917

VL - 1

SP - 247

EP - 265

BT - Probl Control Inf Theory

ER -