Construction of a learning automaton for cycle detection in noisy data sequences

Ustimov A., Tumer B.

COMPUTER AND INFORMATION SCIENCES - ISCIS 2005, PROCEEDINGS, cilt.3733, ss.543-552, 2005 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 3733
  • Basım Tarihi: 2005
  • Dergi Adı: COMPUTER AND INFORMATION SCIENCES - ISCIS 2005, PROCEEDINGS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED)
  • Sayfa Sayıları: ss.543-552
  • Marmara Üniversitesi Adresli: Hayır

Özet

This paper investigates the problem of cycle detection in periodic noisy data sequences. Our approach is based on reinforcement learning principles. A constructive approach is used to devise a variable structure learning automaton (VSLA) that becomes capable of recognizing the potential cycles of the noisy input sequence. The constructive approach allows for VSLAs to analyze sequences not requiring a priori information about their cycle and noise. Consecutive tokens of the input sequence are presented to VSLA, one at a time, where VSLA uses data's syntactic property to construct itself from a single state at the beginning to a topology that is able to recognize an unknown cycle of the given data. The main strength of this approach is applicability in many fields and high recognition rates.