Salvato in:
Dettagli Bibliografici
Autori principali: Qiu, Tianyu, Xie, Yi, Xiong, Yun, Niu, Hao, Gao, Xiaofeng
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2412.15315
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913620510638080
author Qiu, Tianyu
Xie, Yi
Xiong, Yun
Niu, Hao
Gao, Xiaofeng
author_facet Qiu, Tianyu
Xie, Yi
Xiong, Yun
Niu, Hao
Gao, Xiaofeng
contents This paper explores how to enhance existing masked time-series modeling by randomly dropping sub-sequence level patches of time series. On this basis, a simple yet effective method named DropPatch is proposed, which has two remarkable advantages: 1) It improves the pre-training efficiency by a square-level advantage; 2) It provides additional advantages for modeling in scenarios such as in-domain, cross-domain, few-shot learning and cold start. This paper conducts comprehensive experiments to verify the effectiveness of the method and analyze its internal mechanism. Empirically, DropPatch strengthens the attention mechanism, reduces information redundancy and serves as an efficient means of data augmentation. Theoretically, it is proved that DropPatch slows down the rate at which the Transformer representations collapse into the rank-1 linear subspace by randomly dropping patches, thus optimizing the quality of the learned representations
format Preprint
id arxiv_https___arxiv_org_abs_2412_15315
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enhancing Masked Time-Series Modeling via Dropping Patches
Qiu, Tianyu
Xie, Yi
Xiong, Yun
Niu, Hao
Gao, Xiaofeng
Machine Learning
This paper explores how to enhance existing masked time-series modeling by randomly dropping sub-sequence level patches of time series. On this basis, a simple yet effective method named DropPatch is proposed, which has two remarkable advantages: 1) It improves the pre-training efficiency by a square-level advantage; 2) It provides additional advantages for modeling in scenarios such as in-domain, cross-domain, few-shot learning and cold start. This paper conducts comprehensive experiments to verify the effectiveness of the method and analyze its internal mechanism. Empirically, DropPatch strengthens the attention mechanism, reduces information redundancy and serves as an efficient means of data augmentation. Theoretically, it is proved that DropPatch slows down the rate at which the Transformer representations collapse into the rank-1 linear subspace by randomly dropping patches, thus optimizing the quality of the learned representations
title Enhancing Masked Time-Series Modeling via Dropping Patches
topic Machine Learning
url https://arxiv.org/abs/2412.15315