I ran ADMIXTURE with my VCF data

The ADMIXTURE manual says that I can find the appropriate K by checking the cross validation error. what I read about cross validation error is that it is a method of dividing the samples in to folds and train the data with different groups. I'm not even sure what it means in ADMIXTURE cause I couldn't find any specific documents I could understand, and I don't know the algorithms used in ADMIXTURE. I just assume that it uses a model based clustering algorithm and it somehow divides the samples to train the model for clustering(please correct me if I'm wrong). I also don't know how ADMIXTURE calculates the CV error cause there aren't any exact answers.

could someone please briefly explain how ADMIXTURE is done and whether I should use CV error or delta K in order to choose the correct K