thanks..
I found the bug that my Qij in hw7 is wrong.
now I can have optimal value of -123.8
※ 引述《rafan (@_@"""")》之铭言：
: I think there must be something wrong in your program...
: I got obj value -123.272410 on sonar_scale by hw7, and
: -123.272410 by Dr. Lin's program for hw5.
: I think you should think about the whole procedure again. :)
: ※ 引述《mercurial (多变)》之铭言：
: : I run sonar data, and set c=1, r=1/60
: : my hw5 program got -121,
: : and there are many different values for alpha, such as 0, 1, 0.24...
: : however, hw5 doesn't consider the constraint, sum(yi*alpha_i)=0
: : my hw7 program got -193.
: : again, I entered into infinite loop, but I stoped
: : when the optimal value converges to a stable value.
: : however, all the alpha equal to 0 or 1
: : and sum(yi*alpha_i)=0
: : that means the alpha satisify all the constraints
: : I don't understand why adding one more
: : constraint gets better optimal value.
: : maybe my hw5 program is wrong.
: : but my result of hw5 is the same with the professor's.
: : I think I don't fully understand the whole procedure.
: : I just copied down what the professor wrote on board, and
: : implemented them to programs.
: : that's why I couldn't figure out why I got such results.
: : Is is because of bugs or because of
: : the different rules of these two methods?
: : I totally have no idea....