Kernel Networks for Function Approximation

0461978 - ÚI 2017 RIV CH eng C - Conference Paper (international conference)
Coufal, David
Kernel Networks for Function Approximation.
Engineering Applications of Neural Networks. Cham: Springer, 2016 - (Jayne, C.; Iliadis, L.), s. 295-306. Communications in Computer and Information Science, 629. ISBN 978-3-319-44187-0. ISSN 1865-0929.
[EANN 2016. International Conference /17./. Aberdeen (GB), 02.09.2016-05.09.2016]
R&D Projects: GA MŠMT(CZ) LD13002
Institutional support: RVO:67985807
Keywords : kernel networks * convolution * universal approximation
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Capabilities of radial convolution kernel networks to approximate multivariate functions are investigated. A necessary condition for universal approximation property of convolution kernel networks is given. Kernels that satisfy the condition in arbitrary dimension are investigated in terms of their Hankel and Fourier transforms. A computational example is presented to assess approximation capabilities of different convolution kernel networks.
Permanent Link: http://hdl.handle.net/11104/0261515