Title: On Regularity of Max-CSPs and Min-CSPs

Abstract: We study approximability of regular constraint satisfaction problems, i.e., CSPs where each variable in an instance has the same number of occurrences. In particular, we show that for any CSP $\Lambda$, existence of an $\alpha$ approximation algorithm for unweighted regular Max-CSP $\Lambda$ implies existence of an $\alpha-o(1)$ approximation algorithm for weighted Max-CSP $\Lambda$ in which regularity of the instances is not imposed. We also give an analogous result for Min-CSPs, and therefore show that up to arbitrarily small error it is sufficient to conduct the study of approximability of CSPs only on regular unweighted instances.