KDCTime: Knowledge Distillation with Calibration on InceptionTime for Time-series Classification

TL;DR

KDCTime introduces a calibration-based knowledge distillation approach for time-series classification with InceptionTime, significantly improving accuracy and inference speed by addressing overfitting and label calibration issues.

Contribution

The paper proposes KDCTime, a novel method combining calibration and knowledge distillation to enhance InceptionTime for time-series classification, reducing overfitting and inference time.

Findings

KDCTime achieves high accuracy on UCR datasets.

Inference time is two orders of magnitude faster than ROCKET.

The method effectively rectifies teacher model's incorrect soft labels.

Abstract

Time-series classification approaches based on deep neural networks are easy to be overfitting on UCR datasets, which is caused by the few-shot problem of those datasets. Therefore, in order to alleviate the overfitting phenomenon for further improving the accuracy, we first propose Label Smoothing for InceptionTime (LSTime), which adopts the information of soft labels compared to just hard labels. Next, instead of manually adjusting soft labels by LSTime, Knowledge Distillation for InceptionTime (KDTime) is proposed in order to automatically generate soft labels by the teacher model. At last, in order to rectify the incorrect predicted soft labels from the teacher model, Knowledge Distillation with Calibration for InceptionTime (KDCTime) is proposed, where it contains two optional calibrating strategies, i.e. KDC by Translating (KDCT) and KDC by Reordering (KDCR). The experimental results show that the accuracy of KDCTime is promising, while its inference time is two orders of magnitude faster than ROCKET with an acceptable training time overhead.