Bidirectional Generative Pre-training for Improving Healthcare Time-series Representation Learning

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Song, Ziyang, Lu, Qincheng, Zhu, He, Buckeridge, David, Li, Yue
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917757101015040
author Song, Ziyang
Lu, Qincheng
Zhu, He
Buckeridge, David
Li, Yue
author_facet Song, Ziyang
Lu, Qincheng
Zhu, He
Buckeridge, David
Li, Yue
contents Learning time-series representations for discriminative tasks, such as classification and regression, has been a long-standing challenge in the healthcare domain. Current pre-training methods are limited in either unidirectional next-token prediction or randomly masked token prediction. We propose a novel architecture called Bidirectional Timely Generative Pre-trained Transformer (BiTimelyGPT), which pre-trains on biosignals and longitudinal clinical records by both next-token and previous-token prediction in alternating transformer layers. This pre-training task preserves original distribution and data shapes of the time-series. Additionally, the full-rank forward and backward attention matrices exhibit more expressive representation capabilities. Using biosignals and longitudinal clinical records, BiTimelyGPT demonstrates superior performance in predicting neurological functionality, disease diagnosis, and physiological signs. By visualizing the attention heatmap, we observe that the pre-trained BiTimelyGPT can identify discriminative segments from biosignal time-series sequences, even more so after fine-tuning on the task.
format Preprint
id arxiv_https___arxiv_org_abs_2402_09558
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Bidirectional Generative Pre-training for Improving Healthcare Time-series Representation Learning
Song, Ziyang
Lu, Qincheng
Zhu, He
Buckeridge, David
Li, Yue
Artificial Intelligence
Machine Learning
Learning time-series representations for discriminative tasks, such as classification and regression, has been a long-standing challenge in the healthcare domain. Current pre-training methods are limited in either unidirectional next-token prediction or randomly masked token prediction. We propose a novel architecture called Bidirectional Timely Generative Pre-trained Transformer (BiTimelyGPT), which pre-trains on biosignals and longitudinal clinical records by both next-token and previous-token prediction in alternating transformer layers. This pre-training task preserves original distribution and data shapes of the time-series. Additionally, the full-rank forward and backward attention matrices exhibit more expressive representation capabilities. Using biosignals and longitudinal clinical records, BiTimelyGPT demonstrates superior performance in predicting neurological functionality, disease diagnosis, and physiological signs. By visualizing the attention heatmap, we observe that the pre-trained BiTimelyGPT can identify discriminative segments from biosignal time-series sequences, even more so after fine-tuning on the task.
title Bidirectional Generative Pre-training for Improving Healthcare Time-series Representation Learning
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2402.09558