Practical necessity of core-conditioned selector

Determine whether the core-conditioned Togeari-M selector provides a practical advantage over the cheaper raw-factor Togeari-R selector across distinct physical regimes.

Background

Togeari-M selects contact combinations using the core-response Gram matrix UM1UTUM^{-1}U^T, whereas Togeari-R uses the cheaper Euclidean Gram matrix UUTUU^T. In the two tested contact-active extractions, both selectors produced the same five-vector Krylov basis and their retained subspaces were closely aligned.

The core-conditioned selector has the stronger conditional SPD theory, but the experiments do not establish that this theoretical advantage translates into better practical performance. The authors therefore leave its practical value unresolved across broader regimes.

References

In this regime, raw factor geometry recovers it as effectively as core-conditioned selection, so the latter's practical advantage remains unresolved.

TOGEARI: Interaction-Space Preconditioning for Condensed Finite-Element Systems with IPC Contact  (2608.14162 - Liu et al., 14 Aug 2026) in Section 1, paragraph beginning “The distinction between these selectors”

These values quantify close alignment and are consistent with the common five-vector result. Its cause remains unresolved: raw factor geometry already identifies a useful space in the measured regime, while the metric selector alone carries the a priori SPD certificate.

TOGEARI: Interaction-Space Preconditioning for Condensed Finite-Element Systems with IPC Contact  (2608.14162 - Liu et al., 14 Aug 2026) in Section 5.2, paragraph following the principal-angle results