This article presents a new approach based on MiniRocket, called SelF-Rocket, for fast time series classification (TSC). Unlike existing approaches based on random convolution kernels, it dynamically selects the best couple of input representations and pooling operator during the training process. SelF-Rocket achieves state-of-the-art accuracy on the University of California Riverside (UCR) TSC benchmark datasets.
@article{arxiv.2409.01115,
title = {Time series classification with random convolution kernels: pooling operators and input representations matter},
author = {Mouhamadou Mansour Lo and Gildas Morvan and Mathieu Rossi and Fabrice Morganti and David Mercier},
journal= {arXiv preprint arXiv:2409.01115},
year = {2026}
}
Comments
v1: initial version, incorrect evaluation. v2: Method improved, evaluation corrected, title simplified. v3: Add acknowledgments. v4: text correction