Problems involving outer products of vector variables can be relaxed into semidefinite programs. That’s a general trick. Then the low rank bit from SVD is an approixmate solution for the vector

convex relaxation for distributed optimal control

graph matching in relation to Image correspondence

Permutation matrices have sum of rows and columns must be 1 constraint, is one relaxation.

quickMatch. Actually, not convex programming but was the root of the chain of references I ‘m digging through


Finding MaxCut approximation of a graph is a classic one

Quantum Semidefinite programming course

Density matrices have a semidefinite constrina (non negative probabilities)

Sum of Squares is a semidefinite program that can guarantee that lyapunov functions actually work