Robust control methods for on-line statistical learning

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

The issue of controlling that data processing in an experiment results not affected by the presence of outliers is relevant for statistical control and learning studies. Learning schemes should thus be tested for their capacity of handling outliers in the observed training set so to achieve reliable estimates with respect to the crucial bias and variance aspects. We describe possible ways of endowing neural networks with statistically robust properties by defining feasible error criteria. It is convenient to cast neural nets in state space representations and apply both Kalman filter and stochastic approximation procedures in order to suggest statistically robustified solutions for on-line learning.

Original languageEnglish (US)
Pages (from-to)121-127
Number of pages7
JournalEurasip Journal on Applied Signal Processing
Volume2001
Issue number2
DOIs
StatePublished - Jun 2001
Externally publishedYes

Keywords

  • Artificial learning
  • Maximum likelihood inference
  • Robustness and efficiency of estimators
  • Statistical control algorithms

ASJC Scopus subject areas

  • Signal Processing
  • Hardware and Architecture
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Robust control methods for on-line statistical learning'. Together they form a unique fingerprint.

Cite this