The unit Fathom is used to measure the depth of sea. Unfathomable is defined as incapable of being fully explored or understood. Balancing Keeping the family of models Trust-building approximation - RegGAN (supervised learning before unsupervised learning); supervising is based on geometric invariants 3. In general, inferences may be sensitive to features of the underlying distribution of values in the population that cannot be addressed by the observed data. Consequently, for good statistical answers, we not only need models that fit observed data, but we also need: (a) flexibility in these models to allow specification of realistic underlying features not adequately addressed by observed data, such as behavior in the extreme tails of the distribution, or (b) questions that are robust for the type of data collected, in the sense that all relevant underlying features of population values are adequately addressed by the observed values. opt-in vs opt-out policy, onus