Dice Question Streamline Icon: https://streamlinehq.com

Incorporating trivial rescaling/relabeling reductions

Develop a reduction or homomorphism framework that captures trivial equivalences induced by rescaling and relabeling of payoffs (such as the equivalence of 3LIN2 and 3LIN2(1,1)) within the algebraic theory of valued PCSPs.

Information Square Streamline Icon: https://streamlinehq.com

Background

The paper exhibits examples (e.g., 3LIN2 vs. 3LIN2(1,1)) where problems are trivially equivalent but not explained by the current valued minion homomorphism framework. Incorporating these trivial transformations would strengthen the reduction theorem and close gaps between practical reductions and the algebraic theory.

Resolving this would ensure the framework uniformly captures common transformations used in approximation and optimization CSPs.

References

There are however many basic questions and theory-building tasks left open already for finite-domain valued PCSPs: to incorporate the trivial reduction as in e.g. \cref{ex:crips-vs-valued}; to characterize gadget reductions (or versions of definability) in terms of polymorphisms or plurimorphisms; to characterize plurimorphism valued minions of templates; to clarify whether plurimorphisms are necessary to determine computational complexity or enough information is provided already by polymorphisms; to develop methods for proving nonexistence of homomorphisms; to revisit the valued CSP dichotomy without fixed threshold and Raghavendra's result on unique games hardness of approximation for all MaxCSPs; among others. An interesting special case for a full complexity classification is the valued non-promise CSPs with fixed threshold.

The Rise of Plurimorphisms: Algebraic Approach to Approximation (2401.15186 - Barto et al., 26 Jan 2024) in Conclusion