Learning and generalization errors for the 2D binary perceptron

A. Klymovskiy

    Research output: Contribution to journalArticleAcademicpeer-review

    Abstract

    The statistical mechanics model of the binary perceptron learning is considered. It is proved that under the regularity conditions learning and generalization errors for the binary perceptron with two inputs tend to 0 at the average; the first term of the asymptotics is provided; its behavior with respect to the inverse temperature ß does not coincide with that for the model with a smooth activation function.
    Original languageEnglish
    Pages (from-to)1339-1358
    Number of pages20
    JournalMathematical and Computer Modelling
    Volume42
    Issue number11-12
    DOIs
    Publication statusPublished - 2005

    Fingerprint Dive into the research topics of 'Learning and generalization errors for the 2D binary perceptron'. Together they form a unique fingerprint.

    Cite this