Interpolating many matrices without for loop

Hi everybody, well I want to interpolate a 3D matrix composed of 700x2x60 values. I want to interpolate with respect to an array of 700 values. But i want an interpolation only 1 dimension, for example

if i had:

3  1
6  2
9  3

60 times, with different values, i want to interpolate with respect to 1.5, 2.5 and 3 I would get 4.5 7.5 and 10.5, 60 times and with the values relative to each one of the 60 matrices.

So in essence its like interp1 but 60 times. And I can't manage to do it without a for loop which is time consuming. and i need to do it over 15000 times.

thank you for your answers

[Fangjun Jiang]

Without proper {}Code format, it's actually very hard to understand your question.

[j_solar]

the format has been changed a bit. The 3 1 6 2 9 3 is a 3x2 matrix where 3 1 6 is the first column and 2 9 3 the second one.

I have also tried to repmat the array 60 times and use interp2, however it doesn't seem to work right, or I am doing wrong. But I don't think that interp2 is what i need.

thanks again

[j_solar]

how can I put the question in proper format?

[Fangjun Jiang]

You can edit your question and apply {}Code format to make it look right. I did only the {}Code formatting part but didn't change any of your text.

[Fangjun Jiang]

Look at all the buttons above your question text box when you post or edit your question.

Answer 1

[Fangjun Jiang]

You can use reshape().

x=1:3;
y=3:3:9;
X=magic(3);

[M,N]=size(X);
Y=interp1(x,y,X(:),'linear','extrap');
Y=reshape(Y,M,N);

[j_solar]

Ty for the format info, Fangjun.

I'm not sure if my question is clear. Right now I have this:

initial = matrix(700x2x60)

I separate into

a = matrix(700x60);

b = matrix(700x60);

and build an array with values similar to b but not exactly

c = array(700x1);

c = repmat(c,700,1,60)

final = zeros(700x2x60)

final(:,2,:) = c;

for i = 1:60

final(:,1,i) = interp1(a(:,i),b(:,i),c(i));

end

and this works, however I don't like the for loop.

Maybe your answer is right and if so I don't seem to understand

ty

[Fangjun Jiang]

I understand your algorithm now. I don't see an obvious way to avoid the for-loop. arrayfun() might work but in my opinion, that just hides the for-loop.

Why so obsessed with no loops? Numerous examples have shown that no loop doesn't necessarily mean faster.

[j_solar]

I am building a very large simulation program which needs to do many iterations > 15000, (i can't avoid that for loop), with a lot of processing in each iteration. I had bad software(many for loops) and each iteration was taking many seconds(even a minute) which means a veeeery long time. I have managed to reduce time of each iteration to a few seconds, aprox, and i would like to bring it down even more, so I am trying to get rid of all for loops.

But it is true that I have taken some away which have barely reduced my time, and probably it is not worth it eliminating all, but any I can, I try.

thanks once again for everything......... see u in another for loop

[Jan]

Matlab's INTERP1 is lame and there is no reason to limit the algorithm to vectors. But a linear interpolation is not complicated - roughly:

Si = Ti - floor(Ti);

Ti = floor(Ti);

Yi = Y(Ti, :) .* (1 - Si) + Y(Ti + 1, :) .* Si;

I do not understand the value of c. It is a [700x1x60] array any you use c(i) for the interpolation.

[j_solar]

Really c could be a 700x1 array, but since it is going to be the second column of final I replicate it 60 times so I can insert it in final.

Then I use only one(really it could be c(1) instead of c(i)) for the interpolation