论文信息 - GENERATION OF LEARNING NEIGHBOURHOOD IN KOHONEN FEATURE MAPS BY MEANS OF SIMPLE NONLINEAR NETWORK

GENERATION OF LEARNING NEIGHBOURHOOD IN KOHONEN FEATURE MAPS BY MEANS OF SIMPLE NONLINEAR NETWORK

A simple CMOS analogue nonlinear network is presented which can be used as a very efficient means of defining the learning neighbourhood in analogue or digital VLSI realisations of ICohonen maps. The number of components is minimal and the detection of the neighbourhood is facilitated by a steep voltage drop at its boundary.

B. Hochet | Eric A. Vittoz | P. Heim

[1] Teuvo Kohonen,et al. Self-Organization and Associative Memory , 1988 .

[2] Carver Mead,et al. Analog VLSI and neural systems , 1989 .

[3] Teuvo Kohonen. Self-Organizing Feature Maps , 1988 .

[4] Christian-Charles Enz,et al. High precision CMOS micropower amplifiers , 1989 .