SPLD polynomial optimization and bounded degree SOS hierarchies

Abstract: In this paper, a new class of structured polynomials, which we dub the {\it separable plus lower degree {\rm (SPLD in short)} polynomials}, is introduced. The formal definition of an SPLD polynomial, which extends the concept of the SPQ polynomial (Ahmadi et al. in Math Oper Res 48:1316--1343, 2023), is defined. A type of bounded degree SOS hierarchy (BSOS-SPLD) is proposed to efficiently solve the optimization problems with SPLD polynomials, and several numerical examples are performed much better than the bounded degree SOS hierarchy (Lasserre et al. in EURO J Comput Optim 5:87--117, 2017). An exact SOS relaxation for a class of convex SPLD polynomial optimization problems is proposed. Finally, an application of SPLD polynomials to polynomial regression problems in statistics is presented.