Using semidefinite programming to find a better taylor expansion in your newton step sounds like crazy talk to me. No way this could ever be practical. Remembe

Using semidefinite programming to find a better taylor expansion in your newton step sounds like crazy talk to me. No way this could ever be practical. Remember that semidefinite programming is often implemented via the interior point method, which is an application of Newton's method with inequality constraints.