Optimization theory and method study notes
I. Introduction
1 norm
Frobenius norm:
Weighted Frobenius norm weighted L 2 norm (where M is a positive definite symmetric nxn matrix):
Oval vector norm:
In particular, we have
Several important inequalities on norm are:
2, unconstrained problem of optimality conditions
3, a structure optimization method
Second, the one-dimensional search
Third, Newton method
1, steepest descent (gradient descent method, referred to as gradient) - P118
Convergence: linear convergence
2, two step gradient - P127
or
among them,
Convergence: R- superlinear convergence