Signals, Concepts, and Laws: Toward Universal, Explainable Time-Series Forecasting
This addresses the problem of domain-generalizable and interpretable time-series forecasting for applications requiring physical credibility, though it appears incremental as it builds on transformer architectures with added regularization.
The paper tackles the challenge of accurate, explainable, and physically credible forecasting for multivariate time-series with varying statistical properties across domains, proposing DORIC, a transformer-based model that generates predictions through five self-supervised, domain-agnostic concepts while enforcing differentiable residuals grounded in first-principles constraints.
Accurate, explainable and physically credible forecasting remains a persistent challenge for multivariate time-series whose statistical properties vary across domains. We propose DORIC, a Domain-Universal, ODE-Regularized, Interpretable-Concept Transformer for Time-Series Forecasting that generates predictions through five self-supervised, domain-agnostic concepts while enforcing differentiable residuals grounded in first-principles constraints.