Let those two partial derivatives equal to zero.

Edited by author 24.12.2015 20:50

Edited by author 24.12.2015 20:50

is it a some especial test?

:)
Thank you!

I've used 3 different ways to calculate simple linear regression and got 3 different wrong results. Please help.
My email: Vitaly.Arbuzov@gmail.com

i used nested ternary search on a and d but wa ( 24 when eps=1e-7 && 11 when eps<1e-7)
i figured the function should be like that-
f(a,d)=c1a^2+c2a+c3ad+c4d^2+c5d+c6
where c1=n && c4=n*(n-1)*(2*n-1)/6.0
here for n>1 the coefficient of a^2 and d^2 is always positive. so there should be only one(since quadratic) and low peak. so i thought ternary search would do. but why diff wa on diff value of eps. again, i found (by checking for tle) that there are high peaks as well. how can it be? perhaps, i am wrong right on the idea itself. please, guys help me find my bug.
and, please tell me at least the method of solving. is it ternary search or some direct math formula or something else i am missing. if ternary search please tell me whether there is indeed high peaks and how did u deal with them...

thanks in advance.

Edited by author 21.03.2011 17:01

Why in test 1 answer: 0.400 4.900 ?
If b = 0.400, d = 4.900, ∑(ai − bi)^2 =  1,524.
If b = 0.105263, d = 5.026316, ∑(ai − bi)^2 =  0,824 < 1,524.