Renormalization group connection of the multigrid object layer

Investigate whether the coarse-graining performed by the multigrid object layer in Metriplector corresponds to the full renormalization group structure, extending beyond its observed resemblance to block-spin methods.

Background

The multigrid object layer performs learned soft assignment and coarse-level Poisson solves, functionally similar to classical multigrid V-cycles and conceptually akin to block-spin coarse-graining. Clarifying whether this similarity extends to a full renormalization group structure would connect the learned hierarchy to established physics formalisms, potentially yielding principled guarantees or new design constraints.

References

The multigrid object layer's coarse-graining bears a structural resemblance to block-spin methods~\citep{kadanoff1966scaling}; whether this connection extends to the full renormalization group structure is an open question.

— Metriplector: From Field Theory to Neural Architecture  (2603.29496 - Oprisa et al., 31 Mar 2026) in Section 6, Discussion (Limitations and future work)

Until such evidence exists, we do not know which higher-order variables deserve to be kept.

— Is higher-order physics different?  (2609.20759 - Villegas et al., 17 Sep 2026) in Section 7, “What to measure, what to ask, and where to look”