Partial implementation for "Fairness through computationally bounded awareness" by Kim, Reingold, and Rothblum 2018. Written to explore the possibility of post-processing a classifier that was trained on subpopulations. The motivation is using multiple distance arbiters with different domains of expertise.