MATLAB: Vectorization of for/loop

vectorization

I have the following code:

 for ii = 1: length(angles)
    laminate.ply(ii).angle = angles(ii)/180*pi;
    laminate.ply(ii).t = thickness(ii);
    laminate.ply(ii).material = mat_ex3_ge();
 end

How to impliment it without for/loop?

Best Answer

laminate.ply = struct('angles',num2cell(angles(:)/180*pi),...
                      't',num2cell(thickness(:)),...
                      'material',repmat({mat_ex3_ge()},numel(angles),1));

Related Solutions

MATLAB: Increase compute speed compute angle between 2 vectors

The indirection of anonymous functions costs time. So either use the built-in functions with the same names cross, norm and dot, or hard code the functions directly.

Instead of the expensive trick to determine the positions of instabilities in the ASIN and ACOS methods, use a stable method directly:

atan2(norm(cross(N1 x N2)), dot(N1, N2))

Where N1 and N2 are the normalized input vectors.

N1      = bsxfun(@rdivide, a, sqrt(sum(a .* a ,1)))
N2      = bsxfun(@rdivide, b, sqrt(sum(b .* b ,1)))
N1dotN2 = N1(:, 1) .* N2(:, 1) + N1(:, 2) .* N2(:, 2) + N1(:, 3) .* N2(:, 3);
N1xN2   = [(N1(:, 2) .* N2(:, 3) - N1(:, 3) .* N2(:, 2)), ...
         (N1(:, 3) .* N2(:, 1) - N1(:, 1) .* N2(:, 3)), ...
         (N1(:, 1) .* N2(:, 2) - N1(:, 2) .* N2(:, 1))];
Angle   = atan2(sqrt(sum(N1xN2 .* N1xN2, 1)), N1dotN2);

MATLAB: How to transform time series into frequency dimain and obtain amplitude and phase

There's really nothing wrong with the code; you have followed the example from the documentation for FFT and the following lines

    P_amp2=abs(Y/L);
    P_amp(:,jj) = P_amp2(1:L/2+1);
    P_amp(2:end-1,jj) = 2*P_amp(2:end-1,jj);
    f=Fs*(0:(L/2))/L;

return the positive frequency half of the estimated PSD and normalized it to match total power in the double-sided PSD (P_amp2) by the factor of 2 applied to all frequencies except DC and Fmax which are not replicated in the two-sided spectrum.

The frequency vector f is for the one-sided PSD estimate.

If you really want the two-sided PSD, then don't take the positive half elements in P_amp; just use P_amp2 and the corresponding double-sided frequency vector. At that point, for convenience since apparently wouldn't need them, could rename P_amp2 to P_amp and fs to f.

BTW, the FFT and other MATLAB routines are vectorized, you can compute all the results using the original array and dispense with the for loop...

nn_ss=data(1:1800,1:88);           % if size(data)==>1800,88, then this is not needed; use the data array
Ts = mean(diff(time));             % sampling time (by column)
Fs=500;                            %sampling frequency    ** It should be 1/Ts
Fn=Fs/2;                           %Nyquist frequency
L=length(time);
Y=fft(nn_ss);                      % FFT() by column
fts=Y/L;
P_amp=abs(Y/L);                    % 2-sided PSD estimate
phase=angle(Y).*180/pi;
ly=length(Y);
fs=(-ly/2:ly/2-1)/ly*Fs;

The result will be NxM PSD (2-sided) by column same as if had used for...end loop

Best Answer

Related Solutions

MATLAB: Increase compute speed compute angle between 2 vectors

MATLAB: How to transform time series into frequency dimain and obtain amplitude and phase

Related Question