Alternative of REPMAT routine when working with BSXFUN for a large vector

Dear coder, To make thing clear, I will explain the situation with reference to details along with the code below

%%In simple form, the case can be understand as the following

DataSet=5; % 5 data points for each subject
TotalSub=1;
T_sub1=rand((DataSet*TotalSub),1); % For first subject
T_sub2=rand((DataSet*TotalSub),1); % For Second subject

Constant_s=rand(5,2); % Column 1 & 2 represent T1 and T2 respectively

%%Multiplication between each subject to the constant

T_sub1_Constant_s=bsxfun(@times,T_sub1,Constant_s);
T_sub2_Constant_s=bsxfun(@times,T_sub2,Constant_s);

%%In a compact form, we solve similar problem such as
NSub=2; % Assume we have two subjects
T_sub1N2=rand((DataSet*NSub),1); % For first and second subject

Constant_s_rp=repmat(Constant_s,NSub,1); % HOW TO avoid using REPMAT?

T_sub1N2_Constant_s=bsxfun(@times,T_sub1N2,Constant_s_rp);

Yet, as the size of the Constant_s and total subject increase, the vector become extremely large, and using repmat to create these matrices become slower and worse consume to much memory.

I wonder if there is other ways to make the computation fast and memory efficient. I also try KRON, but it does not give me the desired output.

Thanks in advance.

[Jan]

The question would be easier to understand, if you simply provide some test data created by rand(). What are the sizes of a and b? Without your help_repmat.mat file, this is impossible to find out. Which vector becomes "extremely large" and how large is this in absolute numbers?

[balandong]

Dear Jan, Thanks for comment, I already amend the question to a simpler form.

Which vector becomes "extremely large" and how large is this in absolute numbers?

With respect to the above question, the Constant_s.

The size of

DataSet=48; 
TotalSub=1000;
Constant_s=rand(10000,10000);

As you can see, it is very large