核心概念
時系列データにおけるソフトコントラスティブラーニングの効果的な提案とその実験結果を示す。
統計
Contrastive learning has shown to be effective to learn representations from time series in a self-supervised way.
SoftCLT improves the performance in various downstream tasks including classification, semi-supervised learning, transfer learning, and anomaly detection.
SoftCLT improves the average accuracy of UCR datasets and UEA datasets by 2.7% and 3.7%, respectively.