Redundant Hash Addressing of Feature Sequences Using the Self-Organizing Map

  • Authors:

  • Panu Somervuo

  • Affiliations:

  • Neural Networks Research Centre, Helsinki University of Technology, PO Box 2200, FIN-02015 HUT, Finland, e-mail: panu.somervuo@hut.fi

  • Venue:
  • Neural Processing Letters
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Kohonen‘s Self-Organizing Map (SOM) is combined with the Redundant HashAddressing (RHA) principle. The SOM encodes the input feature vector sequenceinto the sequence of best-matching unit (BMU) indices and the RHAprinciple is then used to associate the BMU index sequence with thedictionary items. This provides a fast alternative for dynamicprogramming (DP) based methods for comparing and matching temporalsequences. Experiments include music retrieval and speechrecognition. The separation of the classes canbe improved by error-corrective learning. Comparisons to DP-based methods are presented.