Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Controllable Time Series Generation

Yifan Bao, Yihao Ang, Qiang Huang, Anthony K. H. Tung, Zhiyong Huang

TL;DR

This paper proposes Controllable Time Series Generation (CTSG), an innovative VAE-agnostic framework tailored for CTSG, and extends CTSG to the image domain, highlighting its remarkable potential for explainability and further reinforces its versatility across different modalities.

Abstract

Time Series Generation (TSG) has emerged as a pivotal technique in synthesizing data that accurately mirrors real-world time series, becoming indispensable in numerous applications. Despite significant advancements in TSG, its efficacy frequently hinges on having large training datasets. This dependency presents a substantial challenge in data-scarce scenarios, especially when dealing with rare or unique conditions. To confront these challenges, we explore a new problem of Controllable Time Series Generation (CTSG), aiming to produce synthetic time series that can adapt to various external conditions, thereby tackling the data scarcity issue. In this paper, we propose \textbf{C}ontrollable \textbf{T}ime \textbf{S}eries (\textsf{CTS}), an innovative VAE-agnostic framework tailored for CTSG. A key feature of \textsf{CTS} is that it decouples the mapping process from standard VAE training, enabling precise learning of a complex interplay between latent features and external conditions. Moreover, we develop a comprehensive evaluation scheme for CTSG. Extensive experiments across three real-world time series datasets showcase \textsf{CTS}'s exceptional capabilities in generating high-quality, controllable outputs. This underscores its adeptness in seamlessly integrating latent features with external conditions. Extending \textsf{CTS} to the image domain highlights its remarkable potential for explainability and further reinforces its versatility across different modalities.

Towards Controllable Time Series Generation

TL;DR

This paper proposes Controllable Time Series Generation (CTSG), an innovative VAE-agnostic framework tailored for CTSG, and extends CTSG to the image domain, highlighting its remarkable potential for explainability and further reinforces its versatility across different modalities.

Abstract

Time Series Generation (TSG) has emerged as a pivotal technique in synthesizing data that accurately mirrors real-world time series, becoming indispensable in numerous applications. Despite significant advancements in TSG, its efficacy frequently hinges on having large training datasets. This dependency presents a substantial challenge in data-scarce scenarios, especially when dealing with rare or unique conditions. To confront these challenges, we explore a new problem of Controllable Time Series Generation (CTSG), aiming to produce synthetic time series that can adapt to various external conditions, thereby tackling the data scarcity issue. In this paper, we propose \textbf{C}ontrollable \textbf{T}ime \textbf{S}eries (\textsf{CTS}), an innovative VAE-agnostic framework tailored for CTSG. A key feature of \textsf{CTS} is that it decouples the mapping process from standard VAE training, enabling precise learning of a complex interplay between latent features and external conditions. Moreover, we develop a comprehensive evaluation scheme for CTSG. Extensive experiments across three real-world time series datasets showcase \textsf{CTS}'s exceptional capabilities in generating high-quality, controllable outputs. This underscores its adeptness in seamlessly integrating latent features with external conditions. Extending \textsf{CTS} to the image domain highlights its remarkable potential for explainability and further reinforces its versatility across different modalities.
Paper Structure (56 sections, 4 theorems, 4 equations, 13 figures, 5 tables)

This paper contains 56 sections, 4 theorems, 4 equations, 13 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Our Contributions
  3. Organizations
  4. Related Work
  5. Controllable Image Generation by VAEs
  6. Problem Formulation
  7. The CTS Framework
  8. Overview
  9. Training Phase
  10. Condition Clustering
  11. Generation Phase
  12. Diverse Condition Selection (DCS)
  13. Nearest Neighbor Search (NNS)
  14. Condition Mapping
  15. Theoretical Analysis
  16. ...and 41 more sections

Key Result

Lemma 4.1

For a specific external condition within a valid range, CTS can produce coherent outputs for any altered value $c_0^\prime$ within this range.

Figures (13)

  • Figure 1: The overall pipeline of the CTS framework.
  • Figure 2: An example of Condition Clustering and Data Selection.
  • Figure 3: The pipeline of Condition Mapping.
  • Figure 4: CTSG evaluation scheme.
  • Figure 5: Generation fidelity results on time series datasets.
  • ...and 8 more figures

Theorems & Definitions (10)

  • Definition 3.1: Time Series Generation (TSG) using VAEs
  • Definition 3.2: Controllable Time Series Generation (CTSG)
  • Example 4.1
  • Example 4.2
  • Example 4.3: Interpolation and Extrapolation for Numerical Conditions
  • Example 4.4: Interpolation and Extrapolation for Categorical Conditions
  • Lemma 4.1: Interpolation for Single Condition
  • Corollary 4.1: Interpolation for Multiple Conditions
  • Lemma 4.2: Extrapolation for Single Numerical Condition
  • Corollary 4.2: Extrapolation for Multiple Numerical Conditions