Memory management for data streams subject to concept drift - Sorbonne Université Access content directly
Conference Papers Year : 2016

Memory management for data streams subject to concept drift

Abstract

Learning on data streams subject to concept drifts is a challenging task. A successful algorithm must keep memory consumption constant regardless of the amount of data processed, and at the same time, retain good adaptation and prediction capabilities by effectively selecting which observations should be stored into memory. We claim that, instead of using a temporal window to discard observations with a time stamp criterion, it is better to retain observations that minimize the change in outputted prediction and rule learned with the full memory case. Experimental results for the Droplets algorithm, on 6 artificial and semi-artificial datasets reproducing various types of drifts back this claim.
Fichier principal
Vignette du fichier
Memory management for data streams subject to concept drift - Final.pdf (517.73 Ko) Télécharger le fichier
Origin Files produced by the author(s)
Loading...

Dates and versions

hal-01340498 , version 1 (01-07-2016)

Identifiers

  • HAL Id : hal-01340498 , version 1

Cite

Pierre-Xavier Loeffel, Christophe Marsala, Marcin Detyniecki. Memory management for data streams subject to concept drift. European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Apr 2016, Bruges, Belgium. ⟨hal-01340498⟩
173 View
84 Download

Share

Gmail Mastodon Facebook X LinkedIn More