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

TL;DR

This paper introduces NC-VQVAE, a framework that combines low and high-level semantics in time series latent representations, significantly improving the quality of generated time series data.

Contribution

It proposes a novel integration of self-supervised learning into vector quantization-based time series generation to capture richer semantic information.

Findings

Improved quality of synthetic time series samples.

Effective capture of both low and high-level semantics.

Enhanced modeling of complex time series distributions.

Abstract

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.