How can I optimize this code for vector normalization?

Hi, I implemented this code in order to do the vector normalization of raman intensities. It works, but it is very slow.

%% Vector normalization
% PASSO 1: Calcolare i valori di intensità media per tutti i numeri d'onda
% Le intensità sono salvate in Spectra2
% I ramanshift della fingerprint sono 420
int_FP = load("int_FP.csv");
k = length(int_FP);
nn= size(int_FP);
n= nn(2);
am = zeros(1,n);
corr = zeros(k,1);

for i=1:n
    for ii=1:k
        corr(ii) = int_FP(ii,i);
        
    end
    am(i) = sum(corr);
    
am(i) = am(i)/k;
end
% PASSO 2: Sottraggo alle intensità l'intensità media
clear i ii;
a_k = zeros(k,n);
for ii=1:n
    for i=1:k
    a_k(i,ii) = int_FP(i,ii) - am(ii)
end

end
% PASSO 3:
% Implementare la formula aa_k = a_k/(rad(sum(a_k)^2))
clear i ii corr;
corr = zeros(k,1);
den = zeros(1,n);
for i=1:n
    for ii=1:k
        corr(ii) = a_k(ii,i);
        
    end
    den(i) = sqrt(sum((corr).^2));  

end
clear i ii;
aa_k = zeros(k,n);
for ii=1:n
    for i=1:k
    aa_k(i,ii) = (a_k(i,ii))./den(ii);
end

end

% Verifica algoritmo
clear i ii;
corr = zeros(k,1);
leng = zeros(1,n);
for i=1:n
    for ii=1:k
        corr(ii) = aa_k(ii,i);
        
    end
    leng(i) = sum((corr).^2);
    % leng è 1 per ogni spettro!
  
end
int_FP = aa_k;

How can I optimize this code?

Thanks to anyone who wants to help me.

Answer 1

[albara]

I can see that you're using nested for loops to perform various calculations, which can slow down your code significantly. In MATLAB, it's often faster to use vectorized operations instead of loops. Here's an optimized version of your code:

%% Vector normalization

% PASSO 1: Calcolare i valori di intensità media per tutti i numeri d'onda

% Le intensità sono salvate in Spectra2

% I ramanshift della fingerprint sono 420

int_FP = load("int_FP.csv");

[k, n] = size(int_FP);

% Calculate the mean intensity

am = sum(int_FP, 1) ./ k;

% PASSO 2: Sottraggo alle intensità l'intensità media

a_k = int_FP - repmat(am, k, 1);

% PASSO 3:

% Implementare la formula aa_k = a_k/(rad(sum(a_k)^2))

den = sqrt(sum(a_k.^2, 1));

aa_k = a_k ./ repmat(den, k, 1);

% Verifica algoritmo

leng = sum(aa_k.^2, 1);

int_FP = aa_k;

This optimized code should be much faster. I've replaced the nested loops with vectorized operations, making use of functions like sum, repmat, and element-wise operations. The optimized code should give you the same results as the original code but with improved performance.

Important: There may be some mistakes in this answer Experts can tell if there are any mistakes and you might find better answer

Kindly fgive your feedback

[Valeria Iazzetta]

It works great this way and I will make use of this procedure in the future. Thanks so much for helping me.

[albara]

You are welcome