在这个例子中，我们尝试找到一个Y，使得对于一对随机选择的A（12 \times 12）和B（4 \times 4），\vert\vert AY-YB\vert\vert _F最小化。
>> !cp examples/procrustes/*.m .
>> randn('state',0);
>> [A,B] = randprob;
>> parameters(A,B);
>> Y0 = guess;
>> [fn,Yn] = sg_min(Y0,'newton','euclidean');
iter grad F(Y) flops step type
0 2.334988e+01 3.071299e+01 4751 none
invdgrad: Hessian not positive definite, CG terminating early
1 1.171339e+01 1.376463e+01 365678 Newton step
invdgrad: Hessian not positive definite, CG terminating early
2 7.843279e+00 7.616381e+00 677599 Newton step
invdgrad: Hessian not positive definite, CG terminating early
3 5.131680e+00 4.945824e+00 992823 Newton step
invdgrad: Hessian not positive definite, CG terminating early
4 5.642834e+00 3.512826e+00 1293761 Newton step
invdgrad: Hessian not positive definite, CG terminating early
5 5.500553e+00 1.721329e+00 1607969 Newton step
invdgrad: Hessian not positive definite, CG terminating early
6 4.666307e+00 1.192561e+00 1964675 Newton step
invdgrad: Hessian not positive definite, CG terminating early
7 3.576069e+00 6.850532e-01 2272355 Newton step
invdgrad: max iterations reached inverting the Hessian by CG
8 1.228119e+00 3.046816e-01 2820625 Newton step
9 4.673779e-02 2.506848e-01 3345266 Newton step
10 6.411668e-04 2.505253e-01 3873582 Newton step
11 1.965463e-06 2.505253e-01 4430245 Newton step
12 1.620267e-06 2.505253e-01 5022629 Newton step
图9.1展示了此次运行的收敛曲线。