Abstract
We describe a version of a randomization technique within a general framework of Euclidean Jordan algebras. It is shown how to use this technique to evaluate the quality of symmetric relaxations for several non-convex optimization problems.
Acknowledgement
This work was supported in part by NSF grants DMS04-0270 and DMS07-12809.