Learning, Conjoint Analysis, and Binary Quadratic Optimization
In this talk, we present a learning approach to find good solutions for Binary Quadratic Programming. The proposed approach is based on learning a linear objective function which can then be used to optimize a linear binary program that provides a good feasible solution for the binary quadratic program and is computationally cheaper. The learning approach is inspired from conjoint analysis which can be formulated as a convex quadratic program. Computational results comparing the proposed approach to solving binary quadratic programs using CPLEX are presented.