Some Implications of Interval Approach to Dimension for Network Complexity

0510711 - ÚI 2020 ES eng A - Abstract
Kůrková, Věra
Some Implications of Interval Approach to Dimension for Network Complexity.
ESCIM 2019. Book of Abstracts. Cádiu: University of Cádiz, 2019 - (Kóczy, L.; Medina, J.). s. 59-60. ISBN 978-84-09-14600-0.
[ESCIM 2019: European Symposium on Computational Intelligence and Mathematics /11./. 02.10.2019-05.10.2019, Toledo]
R&D Projects: GA ČR(CZ) GA18-23827S
Institutional support: RVO:67985807
Keywords : quasiorthogonal dimension * sparsity of feedforward networks * high-dimensional geometry * concentration of measure * covering numbers
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

An interval approach to the concept of dimension is presented. Implications of exponentially growing quasiorthogonal dimension for estimates of network complexity are analysed. Bounds on correlations of computational tasks represented by high-dimensional vectors are derived. Network complexity is analyzed from the point of view od the concentration of measure phenomen.
Permanent Link: http://hdl.handle.net/11104/0301114