using repmat instead of for loop

I have a 1x24 matrix A and a 2000x24 matrix D

I want to take column 1 in A and multiply it by all the rows in column 1 of D, then take the second column of A and multiply it by all the rows in column 2 of D and do that for all 24 columns without using a for loop.

A coworker suggested:

G = D .* repmat(A(1:channels), size(D,1),1);

Hoping G is a 2000x24 matrix. It is, but its not doing what i want it to. Any suggestions?

Answer 1

[the cyclist]

That should work. But even better is:

 >> G = bsxfun(@times,A,D);

If that does not give the answer you expect, I think you need to be a little bit more detailed about what you want, possibly with a small explicit example.

Answer 2

[Mason]

I guess im not understanding something. When i added the bxnfun it did the same thing. Here is what im trying to do. I did it correctly using for loops but im trying to do it now without the for loops.

here is with for loops:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% clear all clc d = ReadSegy('MMAC1.SGY'); % load segy data created in knudsen channels = size(d,2); % total number of channels samples = size(d,1); % total number of samples in time shift=max(max(d))/3 ; % offset shift used to plot seismic data highest_values = max(d); % Builds matrix for highest amplitude of each channel highest_value = max(highest_values); % Largest amplitude of data set

%%%%%%%%%%%%%%%%%%%%%%% Plotting the data as Traces %%%%%%%%%%%%%%%%%%%%%%

figure(1)

      for i=1:channels

          plot(d(:,i)+shift*(i-1))                      % plots pre-normalized values
          hold on
          view(90,90)

      end

figure(2) 

      normalized_signals = ones(samples(1),channels(1));  % initializes matrix for plotting normalized signals
      normalized_signals2 = ones(samples(1),channels(1)); % initializes matrix for imagesc of normalized signals
      normalized = highest_values / highest_value;        % matrix of percentages each channel peak value is to the max amplitude value

      for i=1:channels

          normalized_signals(:,i) = d(:,i) ./ normalized(:,i);% normalized signals on all 24 channels to be plotted
          normalized_signals2(:,i) = normalized_signals(:,i) / highest_value;% normalized signals on all 24 channels for imagesc
          plot(normalized_signals(:,i)+shift*(i-1));          % plots normalized values shifted
          hold on
          view(90,90)
      end

figure(3)                                                                   % flips origin to top left corner
    imagesc(normalized_signals2);
    colormap(gray)
    colorbar

%%%%%%%%%%%%%%%%%%%end of program %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Here is what i tried after your suggestion:

clear all
clc

d = ReadSegy('MMAC1.SGY');                      % load segy data created in knudsen
channels = size(d,2);                           % total number of channels
samples = size(d,1);                            % total number of samples in time
shift = max(max(d))/3 ;                         % offset shift used to plot seismic data
highest_values = max(d);                        % Builds matrix for highest amplitude of each channel
highest_value = max(highest_values);            % Largest amplitude of data set

%%%%%%%%%%%%%%%%%%%%%%% Plotting the data as Traces %%%%%%%%%%%%%%%%%%%%%%

figure(1)

      Shifted = d + shift * repmat(1:channels, size(d,1),1 );
      plot(Shifted, 'b');% plots pre-normalized values
      view(90,90)

figure(2) 

      fractions_24 = highest_values / highest_value;   % Makes a 1x24 matrix of fractions of max amplitudes

      scaled = bsxfun(@times,d,fractions_24);         % Attempts to take 1st column of fractions_24 and multiply 
                                                      % it through all rows of the first column of d, 
                                                      % and do this for all 24 channels. 

      normalized = scaled / highest_value;            % Drops all amplitudes to a max of one rest scaled to thatone
      shift = max(max(normalized))/3;
      shifted = shift * bsxfun(@times,1:24,normalized);% Shifts all channels to even spacing
      plot(normalized)                                % Plots the normalized waveforms
      view(90,90)

figure(3)                            
    imagesc(normalized);
    colormap(gray)
    colorbar

%%%%%%%%%%%%%%%%%%end of program %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

[the cyclist]

You had errors in both figures 4 and 5. In both cases, you used "shift" instead of "shift-1". In figure 5, you also multiplied by the shift instead of adding. I think you may also have multiplied instead of divided in one place, but I started to get confused, so I basically went back to your figure 1 and 2 code and translated it directly, rather than starting from your figure 4 and 5 code.

I believe the following replicates what you want.

figure(4)

Shifted = bsxfun(@plus,d,shift*(0:channels-1));

plot(Shifted, 'b');% plots pre-normalized values

view(90,90)

figure(5)

normalized = highest_values / highest_value; % matrix of percentages each channel peak value is to the max amplitude value

normalized_signals = bsxfun(@rdivide,d,normalized);% normalized signals on all 24 channels to be plotted

normalized_signals2 = normalized_signals / highest_value;% normalized signals on all 24 channels for imagesc

shifted = bsxfun(@plus,normalized_signals,shift*(0:channels-1))

plot(shifted,'b'); % plots normalized values shifted

view(90,90)

[Jan]

For the "clear all" see: http://www.mathworks.com/matlabcentral/answers/16484-good-programming-practice#answer_22301

Answer 3

[Mason]

Cyclist, thank you very much. This new function is awesome! I'm trying to do my very best staying away from for loops. You guys rock!

[the cyclist]

Glad to help. Next time, rather than adding your own "Answer" as a response, try using the "Comment on this Answer" feature. As this thread currently stands, it is a little bit tricky for you to accept my answer as being helpful, because that help is spread out.