Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations | Read Paper on Bytez