compare supervised, unsupervised, and self-supervised paradigms to optimize organizational ML pipelines