RT Journal Article
T1 Uncertainty Quantification through Dropout in Time Series Prediction by Echo State Networks
A1 Atencia-Ruiz, Miguel Alejandro
A1 Stoean, Ruxandra
A1 Joya-Caparrós, Gonzalo
K1 Previsión, Teoría de la
K1 Análisis de series temporales
AB The application of Echo State Networks to time series prediction has provided notableresults, favoured by their reduced computational cost, since the connection weights require nolearning. There is however a need for general methods that guide the choice of parameters, inparticular the reservoir size and ridge regression coefficient, improve the prediction accuracy, andprovide an assessment of the uncertainty of the estimates. In this paper we propose such a mechanismfor uncertainty quantification based on Monte Carlo dropout, where the output of a subset of reservoirunits is zeroed before the computation of the output. Dropout is only performed at the test stage,since the immediate goal is only the computation of a measure of the goodness of the prediction.Results show that the proposal is a promising method for uncertainty quantification, providing avalue that is either strongly correlated with the prediction error, or reflects the prediction of qualitativefeatures of the time series. This mechanism could eventually be included into the learning algorithmin order to obtain performance enhancements and alleviate the burden of parameter choice.
PB MDPI
YR 2020
FD 2020-08-17
LK https://hdl.handle.net/10630/41430
UL https://hdl.handle.net/10630/41430
LA eng
NO Atencia, M., Stoean, R., & Joya, G. (2020). Uncertainty Quantification through Dropout in Time Series Prediction by Echo State Networks. Mathematics, 8(8), 1374
NO Ministerio de Ciencia e Innovación. Plan Estatal de Investigación Científica y Técnica y de Innovación
DS RIUMA. Repositorio Institucional de la Universidad de Málaga
RD 20 ene 2026