Blending Low and High-Level Semantics of Time Series for Better Masked Time Series Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mathisen, Johan Vik, Lokna, Erlend, Lee, Daesoo, Aune, Erlend
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913485546323968
author Mathisen, Johan Vik
Lokna, Erlend
Lee, Daesoo
Aune, Erlend
author_facet Mathisen, Johan Vik
Lokna, Erlend
Lee, Daesoo
Aune, Erlend
contents State-of-the-art approaches in time series generation (TSG), such as TimeVQVAE, utilize vector quantization-based tokenization to effectively model complex distributions of time series. These approaches first learn to transform time series into a sequence of discrete latent vectors, and then a prior model is learned to model the sequence. The discrete latent vectors, however, only capture low-level semantics (\textit{e.g.,} shapes). We hypothesize that higher-fidelity time series can be generated by training a prior model on more informative discrete latent vectors that contain both low and high-level semantics (\textit{e.g.,} characteristic dynamics). In this paper, we introduce a novel framework, termed NC-VQVAE, to integrate self-supervised learning into those TSG methods to derive a discrete latent space where low and high-level semantics are captured. Our experimental results demonstrate that NC-VQVAE results in a considerable improvement in the quality of synthetic samples.
format Preprint
id arxiv_https___arxiv_org_abs_2408_16613
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Blending Low and High-Level Semantics of Time Series for Better Masked Time Series Generation
Mathisen, Johan Vik
Lokna, Erlend
Lee, Daesoo
Aune, Erlend
Machine Learning
State-of-the-art approaches in time series generation (TSG), such as TimeVQVAE, utilize vector quantization-based tokenization to effectively model complex distributions of time series. These approaches first learn to transform time series into a sequence of discrete latent vectors, and then a prior model is learned to model the sequence. The discrete latent vectors, however, only capture low-level semantics (\textit{e.g.,} shapes). We hypothesize that higher-fidelity time series can be generated by training a prior model on more informative discrete latent vectors that contain both low and high-level semantics (\textit{e.g.,} characteristic dynamics). In this paper, we introduce a novel framework, termed NC-VQVAE, to integrate self-supervised learning into those TSG methods to derive a discrete latent space where low and high-level semantics are captured. Our experimental results demonstrate that NC-VQVAE results in a considerable improvement in the quality of synthetic samples.
title Blending Low and High-Level Semantics of Time Series for Better Masked Time Series Generation
topic Machine Learning
url https://arxiv.org/abs/2408.16613