I have a small multivariate data ( < 10,000 records) with none of the variable following normal distribution (mix of right-skewed and left-skewed variables). As Gaussian Mixture Model is not suitable for this, what are the other methods that allows computing a probability to belong to each cluster, for every point ?