Title :
A method to obtain better codebooks for vector quantizers than those achieved by the generalized Lloyd algorithm
Author :
Yair, Eyal ; Zeger, Kenneth
Author_Institution :
IBM Israel Sci. Center, Technion City, Haifa, Israel
Abstract :
The method is based on a deterministic on-line design technique (similar to the Kohonen learning scheme) which incorporates principles of stochastic relaxation. In contrast to the Kohonen winner-take-all competition between the candidate codevectors, it performs a `soft´ competition where each codevector is assigned a winning score and all codevectors are updated simultaneously according to these scores. A temperature schedule is used to control the speed of convergence versus the quality of the codebook
Keywords :
convergence; intelligent design assistants; learning (artificial intelligence); neural nets; online operation; vector quantisation; Kohonen learning scheme; codebooks; deterministic online design; generalized Lloyd algorithm; quality; soft competition; speed of convergence; stochastic relaxation; temperature schedule; vector quantizers; winning score; Algorithm design and analysis; Cities and towns; Convergence; Design methodology; Iterative algorithms; Least squares approximation; Prototypes; Stochastic processes; Temperature control; Vector quantization;
Conference_Titel :
Electrical and Electronics Engineers in Israel, 1991. Proceedings., 17th Convention of
Conference_Location :
Tel Aviv
Print_ISBN :
0-87942-678-0
DOI :
10.1109/EEIS.1991.217666
