Fitting 65k rows with 6 points takes too much time with for loop.

Dear helpful matlab users,

Previously I've asked a quesition about plotting +60k lines, by the help of you time cost reduced for plotting. But this time my issue is about generating slopes/coefficients.

I have two matrices which one of them has 65536 rows and 6 colums and the other one has 1 row 6 columns.. Let's say those matrices are M_1(1,6) and M_2(65536,6)

I would like to fit all rows as given below code section.

coefficients=zeros (65563,2);
for i=1:65536
    coefficients(i,:) = polyfit (M_1, M_2(i,:),1);
end

Is there a way that I can reduce the time requires to calculate that portion of code.

for loop slows down everything.

Best Regars,

Ömer

Answer 1

[Chunru]

n = 65536;
M_1 = randn(1, 6);
M_2 = randn(n, 6);

tic
coefficients=zeros (n,2);
for i=1:n
    coefficients(i,:) = polyfit (M_1, M_2(i,:),1);
end
toc

Elapsed time is 0.934269 seconds.

% coefficients

% Polyfit cab be done by solving th vandermonde matrix system
% doc polyfit [and check out the algo]
tic
a = M_1'.^flip([0 1]);        % 1st order polynomial
c = (a \ M_2')';
toc

Elapsed time is 0.009172 seconds.

% To check if the results agree
max(abs(c(:) - coefficients(:)))

ans = 4.4409e-16

[Ömer Yaman]

Thank you that is totaly worked.