Multivariate Extension Challenge

Background: The scope of a proof-of-concept study often necessitates simplification of evaluation settings, which may mask issues when scaling to more complex data structures.

Question / Future Work: The current implementation is restricted to univariate time series with fixed window sizes (n=32, m=16). Extending the retrodictive framework to multivariate time series will introduce significant complexity, especially regarding the MAP optimization routine over the augmented future variable 𝒚.