Can anyone help me figure out the preallocation of the unknown matrix which no. of rows changing in every cycle? Thank in advance

close all; clc;
tic
p=dir('*.txt');
nFiles=numel(p);
A_cat=[];  % A_cat=zeros(500000,30);
for k=1:nFiles
    A=dlmread(p(k).name);
    disp(size(A_cat)); %% you can check the no. of rows is increasing
    %A_cat=[A_cat;A]; %% every cycle, matrix changes its size!!!
    A_cat=vertcat(A_cat,A);
end
toc

Answer 1

[KSSV]

close all; clc;
tic
p=dir('*.txt');
nFiles=numel(p);
A_cat=cell(nFiles,1);  % A_cat=zeros(500000,30);
for k=1:nFiles
    A=dlmread(p(k).name);
    disp(size(A_cat)); %% you can check the no. of rows is increasing
    %A_cat=[A_cat;A]; %% every cycle, matrix changes its size!!!
    A_cat{k} = A;
end
A = vertcat(A_cat{:}) ;
toc

[SA]

Many thanks for the response. it works fine and takes only ~353 sec while the other code took ~1200 sec. I'm curious is there any other way to execute the operation more faster or the fastest way? Thanks aga!n

[KSSV]

There is a small typo error..now changed..check the time now...

A_cat=cell(nFiles,1); 

[SA]

I've checked the profile summary, basically 90% time consumed by the 'dlmread' command..while cell size would not make a big difference. Indeed, thanks for your effort.

For the matrix A[5e7, 30], if I want to separate each column and wants to find the mean.....then what could be the faster approach.....any observations....

for j =1:30

P_dat = buffer(A(:,j),500); %% cut the 500 elements*1000 of a single sensor

P_avg(:,j)=mean(P_dat,2);

end

[KSSV]

Try other functions instead of dlmread. Have a look on load, importdata.

[SA]

do you think the other possibility to run the code faster?

tic;

P_avg=cell(1,30);

for j =1:30

P_dat = buffer(A(:,j),500); %% cut the 500 elements*1000 of a single sensor

P_m=mean(P_dat,2);

P_avg{j}=P_m;

end

mn_dat = horzcat(P_avg{:}) ;

toc;