Ordering strategy for applying certified joins

Determine an effective strategy for ordering the application of certified join operations when iteratively combining cut and join partial optimality conditions for cubic correlation clustering, including how the ordering impacts the fraction of variables that can be fixed and how to select heuristics to optimize this effect.

Background

The paper shows that cut and join conditions interact and that joins can change the instance, after which conditions are re-applied iteratively. The authors emphasize that the fraction of variables fixed depends on the order in which certified joins are applied.

They explicitly leave open which ordering strategy should be used, suggesting that heuristic design for ordering may significantly affect practical performance.

References

One question we do not answer here is in which order to apply certified joins. The fraction of fixed variables depends on this order, and looking into heuristics for choosing an order appears promising.

Partial Optimality in Cubic Correlation Clustering for General Graphs (2510.20431 - Stein et al., 23 Oct 2025) in Conclusion