Can Bayesian Neural Networks Explicitly Model Input Uncertainty?
This addresses the problem of ignoring input noise in machine learning models, which is incremental as it tests existing BNN methods for a specific capability.
The paper investigated whether Bayesian Neural Networks (BNNs) and their approximations can model input uncertainty, finding that only Ensembles and Flipout among the tested methods are capable of this.
Inputs to machine learning models can have associated noise or uncertainties, but they are often ignored and not modelled. It is unknown if Bayesian Neural Networks and their approximations are able to consider uncertainty in their inputs. In this paper we build a two input Bayesian Neural Network (mean and standard deviation) and evaluate its capabilities for input uncertainty estimation across different methods like Ensembles, MC-Dropout, and Flipout. Our results indicate that only some uncertainty estimation methods for approximate Bayesian NNs can model input uncertainty, in particular Ensembles and Flipout.