On support vector machines and sparse approximation for random processes

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

After reviewing the role played by support vector machines and other sparse approximation instruments in studies about random processes, this work aims to emphasize some of their properties and proposes a space dimensionality reduction approach which is conceptually simple and computationally feasible. The stochastic analysis of random phenomena is often required in a context characterized by non-standard probabilistic assumptions. In other terms, real-life problems offer statistical model conditions such as strong forms of dependence, non-stationarity, non-gaussianity, for instance; it often appears that these features prevent from using classical statistical inference procedures or model-building strategies, and instead require computational techniques based on simulations and numerical approximation. One initial problem, here investigated, is how to cast these instruments in a certain theoretical framework, for then achieving suitable model structure and interpretation. This contribution, which is of an introductory nature, suggests further advances into the domain of real life applications, in particular time series.

Original languageEnglish (US)
Pages (from-to)39-60
Number of pages22
JournalNeurocomputing
Volume56
Issue number1-4
DOIs
StatePublished - Jan 2004
Externally publishedYes

Keywords

  • Frames
  • Integral equations
  • Random processes
  • Reproducing kernel Hilbert spaces
  • Sparse approximation
  • Support vector machines
  • Wavelets

ASJC Scopus subject areas

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'On support vector machines and sparse approximation for random processes'. Together they form a unique fingerprint.

Cite this