how to resolve the error

[pc,score,latent,tsquare] = princomp(x); red_dim = score(:,1:50);

??? Error using ==> svd Out of memory. Type HELP MEMORY for your options.

Error in ==> princomp at 85 [U,sigma,coeff] = svd(x0,econFlag); % put in 1/sqrt(n-1) later

MY x value consist 50x36033 i need to use feature reduction

Answer 1

[Aditya]

Hi Ajith,

The error you're encountering is due to memory limitations when trying to perform Singular Value Decomposition (SVD) on a large matrix using princomp. Since princomp is deprecated and you have a large dataset, it's better to use pca, which is more efficient and offers better handling of large datasets.

% Assume X is your data matrix with observations in rows and variables in columns
% In your case, X is 50x36033

% Perform PCA
[coeff, score, latent, tsquared, explained, mu] = pca(X, 'NumComponents', 50);

% score contains the reduced dimensions (50x50)
% coeff contains the principal component coefficients
% latent contains the eigenvalues
% tsquared contains Hotelling's T-squared statistic
% explained contains the percentage of total variance explained by each component
% mu contains the estimated mean of each variable

% Extract the reduced dimensions
red_dim = score; % This will be 50x50

% If you need to transform new data into the reduced space, use:
% new_data_transformed = (new_data - mu) * coeff(:, 1:50);

Using pca instead of princomp should help you perform PCA more efficiently on large datasets and reduce the dimensionality to the desired number of components.