matrix and its transpose matrix

I have:

clc;clear;close all;
N = 10;
SNRdB=0:2:20;
SNR = 10.^(SNRdB/10);
P=2;
B = 1000;
I = 3; 
for j=1:length(N)
    A = 0.5.*(randn(3,3, N) +(1i) * randn(3,3, N)); 
end
C = log2(det(I + (P/N)*(A*A')));
figure(1)
semilogy(SNRdB,C_eigen,'ro-','lineWidth',.5);

receive error:

Error using '

Transpose on ND array is not defined. Use PERMUTE instead.

Error in matrix33e (line 12)

C = log2(det(I + (P/N)*(A*A')));

Please help

Thanks

Answer 1

[Walter Roberson]

for j=1:length(N)
    A = 0.5.*(randn(3,3, N) +(1i) * randn(3,3, N)); 
end

Why are you overwriting all of A in each loop iteration?

You are creating A as a 3 x 3 x 10 array each time.

C = log2(det(I + (P/N)*(A*A')));

Reminder: the * operator is only valid when one of the sides is a scalar, or both sides are either vectors or 2D arrays with the number of columns of the first operand being the same as the number of rows of the second operand. Your 3 x 3 x 10 array is not 2D so * cannot be used with it.

N = 10;
SNRdB=0:2:20;
[...]
semilogy(SNRdB,C_eigen,'ro-','lineWidth',.5);

I would suggest to you that you should not be assigning a constant N, and should instead be using

N = length(SNRdB);

I would further suggest to you that you should be using

for j=1:N
    A = 0.5.*(randn(3,3) +(1i) * randn(3,3));
    C = log2(det(I + (P/N)*(A*A')));
    C_eigen(j) = something having to do with C
end

[Bruno Luong]

for j=1:length(N)

That runs only once

[Seahawks]

But I need to set N = 10000 random matrices. Please advice.

[Walter Roberson]

Okay so you want N = 10000 random matrices. How does that connect to your 11 different signal to noise ratios? And when you plot, what size of output are you expecting?

[Seahawks]

My code generate 10000 matrices but could not find transpose H'

[Walter Roberson]

Good catch, Bruno, I have fixed that part.

[Seahawks]

The 10000 of 3x3 matrices. Size to plot 0:2:20

[Walter Roberson]

So for each of the entries in 0:2:20 you want to generate 10000 matrices that are 3 x 3 ? Are you expecting your output to be length(0:2:20) by 10000, or do you have code that somehow summarizes the results for each of the 10000 3 x 3 matrices into a single value?

For example are you finding the maximum absolute eigenvalue for each C, and your summary over the 10000 matrices is to find the largest absolute value out of the 10000 tries? To, in other words, summarize what the "worst case" (best case?) is?

[Seahawks]

Each generate 10000 matrix to 10000 values. These 10000 values will go into C formula to plot with SNRdB. Hope it clear for you

[Seahawks]

I found the way to do that. Thanks