Number of the records: 1
Recent Trends in Learning from Data
- 1.
SYSNO ASEP 0521198 Document Type M - Monograph Chapter R&D Document Type Monograph Chapter Title Limitations of Shallow Networks Author(s) Kůrková, Věra (UIVT-O) RID, SAI, ORCID Source Title Recent Trends in Learning from Data. - Cham : Springer, 2020 / Oneto L. ; Navarin N. ; Sperduti A. ; Anguita D. - ISSN 1860-949X - ISBN 978-3-030-43882-1 Pages s. 129-154 Number of pages 26 s. Number of pages 221 Publication form Print - P Language eng - English Country CH - Switzerland Keywords shallow and deep networks ; model complexity ; probabilistic lower bounds Subject RIV IN - Informatics, Computer Science OECD category Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8) R&D Projects GA18-23827S GA ČR - Czech Science Foundation (CSF) Institutional support UIVT-O - RVO:67985807 EID SCOPUS 85085176004 DOI 10.1007/978-3-030-43883-8_6 Annotation Although originally biologically inspired neural networks were introduced as multilayer computational models, shallow networks have been dominant in applications till the recent renewal of interest in deep architectures. Experimental evidence and successfull application of deep networks pose theoretical questions asking: When and why are deep networks better than shallow ones? This chapter presents some probabilistic and constructive results on limitations of shallow networks. It shows implications of geometrical properties of high-dimensional spaces for probabilistic lower bounds on network complexity. The bounds depend on covering numbers of dictionaries of computational units and sizes of domains of functions to be computed. Probabilistic results are complemented by constructive ones built using Hadamard matrices and pseudo-noise sequences. Workplace Institute of Computer Science Contact Tereza Šírová, sirova@cs.cas.cz, Tel.: 266 053 800 Year of Publishing 2021
Number of the records: 1