DocumentCode :
671733
Title :
DPLL based hardware SOM with a new winner-take-all circuit
Author :
Hikawa, Hiroomi
Author_Institution :
Sci. & Eng., Kansai Univ., Suita, Japan
fYear :
2013
fDate :
4-9 Aug. 2013
Firstpage :
1
Lastpage :
8
Abstract :
The self-organizing map (SOM) has found applicability in a wide range of areas. This paper proposes a new winner-take-all method for hardware SOM with phase-modulated pulse signal. The system uses phase of square waveform to hold the value of each input vector element. The system uses digital phase-locked loop (DPLL) as a computing element because the operation of the DPLL is very similar to that of SOM´s computation. The proposed winner search method is realized by taking advantage of the nature of phase modulated signal, and neighborhood function is also implemented with simple hardware. In addition, the winner search requires no global communication, which makes the architecture scalable. The proposed SOM architecture is described in VHDL and its feasibility is verified by simulation. Preliminary simulation results show that the proposed SOM has good clustering capability with the topology preserving nature.
Keywords :
digital phase locked loops; hardware description languages; neural net architecture; phase modulation; search problems; self-organising feature maps; DPLL based hardware SOM; SOM architecture; VHDL; clustering capability; digital phase-locked loop; input vector element; neighborhood function; phase-modulated pulse signal; self-organizing map; winner search method; winner-take-all circuit; winner-take-all method; Clocks; Computational modeling; Computer architecture; Hardware; Neurons; Radiation detectors; Vectors; DPLL; VHDL; phase modulated signal; self organizing map;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Neural Networks (IJCNN), The 2013 International Joint Conference on
Conference_Location :
Dallas, TX
ISSN :
2161-4393
Print_ISBN :
978-1-4673-6128-6
Type :
conf
DOI :
10.1109/IJCNN.2013.6707075
Filename :
6707075
Link To Document :
بازگشت