scientific article; zbMATH DE number 3978844
A new concise proof of the following theorem is given. Theorem. Consider the unconstrained problem \[ \min f(x)=1/2 x^ TGx+r^ Tx+\delta,\quad x\in R^ n \] where G is a positive symmetric matrix. Suppose \(x^*\) denotes the minimimum of f and \(x_ n=x_ n(x_ 0;D)\) denotes the minimum of f after searching from \(x_ 0\) along \(D=(d_ 1,...,d_ n)\). We then have \[ (*)\quad f(x_ n(x_ 0;D))- f(x^*)\leq (1-\mu^ 2_ n)[f(x_ 0)=f(x^*)] \] where \(\mu_ n\) is the measure of conjugacy of the direction set D \[ \mu_ n=0,\;if\;\prod^{n}_{i=1}\| d_ i\| =0; \] \[ \mu_ n=| \det \sqrt{G}| \cdot | \det \{d_ 1,...,d_ n\}| /\prod^{n}_{i=1}<d_ i,Gd_ i>^{1/2},\;if\;\prod^{n}_{i=1}\| d_ i\| \neq 0. \] Furthermore the equality in (*) can be obtained.
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3743124)