Prima pagina Indietro Avanti Ultima pagina Panoramica Immagine

The typical Modern Spam Filter

  • Bayesian classifier
  • Training:
  • Train On Errors (TOE strategy)
  • Train Every Thing (TEFT strategy, a.k.a. “bulk” training)
  • Train Until No Errors (TUNE strategy)
  • Only top N features (or “peaks”) are used for classification
  • Is this Optimal?

    Note: