Multivariate Stable Distributions and Temporal Tail Dynamics
Background: Lévy stable distributions are used to model heavy-tailed processes but lack closed-form probability density functions, which makes conventional maximum likelihood estimation difficult. Deep learning models typically rely on parametric distributions with closed-form densities to facilitate gradient-based optimization for probabilistic forecasting.
Question / Future Work: Extending the framework to multivariate Lévy stable distributions and modeling the dynamic temporal dependency of tail parameters within sequence-to-sequence architectures remains an open challenge for improved extreme event forecasting.
Why It Matters: Multivariate modeling is essential for capturing inter-asset correlations during market shocks, and temporal tail dynamics would allow for more accurate dynamic risk assessment in non-stationary series.
Evidence: Future work will explore multivariate stable distributions and modeling temporal dependence in tail parameters.
Metadata & Links
- created_at
- 2026-05-14T06:08:57Z