Time series classification with random convolution kernels: pooling operators and input representations matter

TL;DR

SelF-Rocket is a novel, fast time series classifier that dynamically selects optimal input representations and pooling operators, achieving state-of-the-art accuracy on benchmark datasets.

Contribution

It introduces a dynamic selection mechanism in MiniRocket, improving accuracy without sacrificing speed in time series classification.

Findings

Achieves state-of-the-art accuracy on UCR benchmark datasets.

Dynamically selects input representations and pooling operators during training.

Abstract

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.