I am trying to do a backtracking line search algorithm for a stepest decent problem. This bascial means I need to keep using the gradiant and eventually the hessian alot. But I am trying to compare the operators to be able to compare and I just cant get it. below is my code and a photo of what i am trying to do.

 채택된 답변

Walter Roberson
Walter Roberson 2020년 12월 4일

0 개 추천

You missed the point that requires that the gradient of f be evaluated at x. You are taking the symbolic gradient of f into gf, but you are not evaluating it at x.
Note: if you have a symbolic gradient gf, then you do not need to re-calculate it every iteration: you can calculate gf once and evaluate gf with the current x at every step.

댓글 수: 5

Chance Anderson
Chance Anderson 2020년 12월 4일
thank you
Chance Anderson
Chance Anderson 2020년 12월 4일
So on this function the orginal function would be comprised of x_1, x_2, x_3, x_4, x_5, x_6 how would one evaluate the overall funtion at one particule point?
subs(gf, [x_1, x_2, x_3, x_4, x_5, x_6], x)
Chance Anderson
Chance Anderson 2020년 12월 4일
hey thanks for the help, im not very good at this and your help is very useful cause.
So im sorry to ask again
I tried using this evalualtion at several different points of the code. Am i using it in the wrong context
f =
-log((x_2/5 - 1)*(x_5/2 - 1)*((2*x_4)/5 - 1)*(x_1/10 - 1)*((3*x_6)/5 - 1)*((3*x_3)/10 - 1)*(x_1^2 - 1)*(x_2^2 - 1)*(x_3^2 - 1)*(x_4^2 - 1)*(x_5^2 - 1)*(x_6^2 - 1))
gf =gradient(f);
subs(gf, [x_1, x_2, x_3, x_4, x_5, x_6], x);
deltax = -1*gf;
a = .3;
b = .8;
t = 1;
inc = 1;
step = 0;
startq = f + a * t * gf.' * deltax;
track = f + a * t * gf.' * deltax;
while startq < track
t = t * B^inc;
disp(inc);
inc = inc +1;
disp(track)
end

댓글을 달려면 로그인하십시오.

추가 답변 (0개)

카테고리

도움말 센터File Exchange에서 Symbolic Math Toolbox에 대해 자세히 알아보기

태그

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by