BayesInference considers three possible types of given discrepancy, each with a different interface. This makes the Inference difficult to read and modify. Can this be improved by moving this distinction into the Discrepancy class?
changed milestone to %2.0.0
added Inference Restructuring labels
marked this issue as related to #57 (closed)
Create flags so the user can determine which errors to consider: a) only measurement error, b) consider RMSE as surrogate error (check if validation data was given and RMSE was estimated, c) consider surrogate error (check if surrogate gives Gaussian distribution over posteriors).
assigned to @RKohlhaas
!47 (merged): Options were removed, now only allow a) known measurement uncertainty.
closed