In a numerically perfect world,

would be exactly 0 except at the support vectors, and so

would give the same result is

.

On real computers, of course, we have to deal with the fact that our calculations have limited precision, and so

is usually non-zero nearly everywhere.

I found that if I identified the support vectors before calculating

, by looking for

for some small

, and then calculated

just from those support vectors, I did not get a consistent

. If there were 3 support vectors, sometime I'd get the same

from all 3, but maybe half the time I'd get one

from two of them, and the third would give a

that was significantly off.

If, however, I used all the vectors to calculate

, rather than just the support vectors, then I'd get the same

from all the support vectors.

My speculation is that just as the

values that are supposed to be 0 are off slightly due to floating point precision issues, so too are those that are supposed to be non-zero, and that when you use ALL of the

's to calculate

the errors are balancing out. When you exclude the ones that were "supposed" to be 0, you increase the error in

. This makes intuitive sense because the QP solver was using all the

's to try to achieve minimization, and so any error should be spread among all of them. If we only have 3 support vectors, and so only use 3

's, the error will be high because 3 is so small we get high variance. By using all the

's, the variance will be lower, and so the error is closer to the mean error, which should be zero.

Those who had errors on problems 8-10, if you just used the support vectors, and calculated

from one support vector, it might be worth putting in a check to see if you get a different

from different support vectors.