Universal Time Series Generation with Neural Controlled Differential Equations
For researchers in time-series modelling, this work provides a theoretically grounded, maximally expressive generative model that handles irregularly sampled data effectively.
The paper proves that Structured Linear Controlled Differential Equations (SLiCEs) are universal time-series generators and proposes Generative SLiCEs (G-SLiCEs) for flow matching on path-space, achieving improved performance in probabilistic forecasting and downstream tasks, especially on irregular grids.
Recent work on the sequence universality of State Space Models (SSMs) has introduced efficient, maximally expressive continuous-time approaches for time-series modelling. While these works focus on discriminative settings, we extend this perspective to generative time-series modelling by proving that maximally expressive Structured Linear Controlled Differential Equations (SLiCEs) are universal time-series generators, in the sense that they can approximate the induced path laws of continuous causal pushforwards on compact latent sets in $W_\infty$. Building on these theoretical results, we propose Generative SLiCEs (G-SLiCEs), a maximally expressive continuous-time model for flow matching on path-space. Empirically, we show that expressivity improves performance in probabilistic forecasting and downstream tasks, while retaining the advantages of continuous-time models such as generalising to arbitrary observation grids. This is particularly beneficial for irregular grids, where fixed-grid models often struggle.