LDA analysis: The pooled covariance matrix of TRAINING must be positive definite.

Hello, I am running into this issue. How can I resolve it?

Y = csvread('mydata.csv');
flag = Y(:,1);
label = Y(:,2);
P = Y(:,3:end);
train = Y((flag < 5) & (label == 8|9),:);
test = Y((flag == 5) & (label == 0),:);

[coeff,score,latent] = pca(train);
group = Y((flag < 5) & (label == 8|9));
class = classify(Y,train,group,'linear');

My research online gives me some hints that I should apply PCA to the training samples and project onto the first 2 principal components. Then, apply LDA to project onto 1 dimension.

How can I take the result of PCA and input it as a parameter in classify()?

Thank you!

Answer 1

[Fadi Alsuhimat]

Just write it like this

augmentedTrainset=augmentedImageDatastore(imagesize,...

trainset,'ColorPreprocessing','gray2rgb');

%%% this mean you add another type for lda by using 'ColorPreprocessing','gray2rgb'