Reference implementation of Robust Probabilistic Modeling with Bayesian Data Reweighting. [arxiv]
This folder contains code to replicate:
-
empirical study on a Poisson model with corrupted observations
-
empirical study on a Generalized Linear model with missing latent groups. The code illustrates the use of Dirichlet prior on weights in the example.
-
empirical study on a Gaussian Mixture model with skewed distributions