A modified Lanczos Algorithm for fast regularization of extreme learning machines

Renjie Hu*, Edward Ratner, David Stewart, Kaj Mikael Björk, Amaury Lendasse

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

Abstract

This paper presents a new regularization for Extreme Learning Machines (ELMs). ELMs are Randomized Neural Networks (RNNs) that are known for their fast training speed and good accuracy. Nevertheless the complexity of ELMs has to be selected, and regularization has to be performed in order to avoid underfitting or overfitting. Therefore, a novel Regularization is proposed using a modified Lanczos Algorithm: Iterative Lanczos Extreme Learning Machine (Lan-ELM). As summarized in the experimental Section, the computational time is on average divided by 4 and the Normalized MSE is on average reduced by 11%. In addition, the proposed method can be intuitively parallelized, which makes it a very valuable tool to analyze huge data sets in real-time.

Original languageEnglish
Peer-reviewed scientific journalNeurocomputing
Volume414
Pages (from-to)172-181
Number of pages10
ISSN0925-2312
DOIs
Publication statusPublished - 13.11.2020
MoE publication typeA1 Journal article - refereed

Keywords

  • Classification
  • Extreme Learning machines
  • Lanczos Algorithm
  • Neural Networks
  • Regression
  • Regularization

Fingerprint Dive into the research topics of 'A modified Lanczos Algorithm for fast regularization of extreme learning machines'. Together they form a unique fingerprint.

  • Cite this