Avoiding overfitting of models: an application to research data on the Internet videos

0481488 - ÚTIA 2018 RIV CZ eng K - Konferenční příspěvek (tuzemská konf.)
Jiroušek, Radim - Krejčová, I.
Avoiding overfitting of models: an application to research data on the Internet videos.
Proceedings of the 35th International Conference Mathematical Methods in Economics (MME 2017). Hradec Králové: University of Hradec Králové, 2017, s. 289-294. ISBN 978-80-7435-678-0.
[MME 2017. International Conference Mathematical Methods in Economics /35./. Hradec Králové (CZ), 13.09.2017-15.09.2017]
Grant ostatní: GA ČR(CZ) GA15-00215S
Institucionální podpora: RVO:67985556
Klíčová slova: data-based learning * probabilistic models * information theory * MDL principle * lossless encoding
Obor OECD: Applied Economics, Econometrics
http://library.utia.cas.cz/separaty/2017/MTR/jirousek-0481488.pdf

The problem of overfitting is studied from the perspective of information theory. In this context, data-based model learning can be viewed as a transformation process, a process transforming the information contained in data into the information represented by a model. The overfitting of a model often occurs when one considers an unnecessarily complex model, which usually means that the considered model contains more information than the original data. Thus, using one of the basic laws of information theory saying that any transformation cannot increase the amount of information, we get the basic restriction laid on models constructed from data: A model is acceptable if it does not contain more information than the input data file.
Trvalý link: http://hdl.handle.net/11104/0277045