Multivariate Optimization and constraints

A few errors I caught: Inverting the Hessian is worst-case O(d^3), but multiplying the inverse Hessian by the gradient is just O(d^2) because it's matrix-vector multiplication