Learning temporal matchings for time series discrimination
Texte intégral
Documents relatifs
The challenge in a database of evolving time series is to provide efficient algorithms and access methods for query processing, taking into consideration the fact that the database
For nearest neighbor time series classication, we propose to learn a metric as a combination of temporal and frequential metrics based on a large margin opti- mization process..
• Data mining domain has techniques for algorithmically examining time series data, looking for patterns, etc.. • Good when objective is known
We adopt two similarity measures, Euclidean Distance (ED) and Dynamic Time Warping (DTW), to calculate the distance between two time series and show the efficacy of our proposed
The aim of this work is to show that the forecasting performance could be improved by replacing the bootstrap step by what we call block bootstrap variants, to subsample time
These methods take advantage of unlabeled data: the behavior of functions corresponding to different choices of complexity are compared on the training data and on the unla- beled
(At the time of such analyses the generating processes should be unknown, but can be checked afterwards.) Even so the situation is better than can be expected with «real"
Leaf area index estimation with MODIS reflectance time series and model inversion during full rotations of Eucalyptus plantations. Relating MODIS vegetation index time-series