Uncertainty Decomposition in Bayesian Neural Networks with Latent Variables

By Stefan Depeweg, José Miguel Hernández-Lobato, Finale Doshi-Velez, Steffen Udluft. 2017

46 citations recorded by the source corpus.

Bayesian neural networks (BNNs) with latent variables are probabilistic models which can automatically identify complex stochastic patterns in the data. We describe and study in these models a decomposition of predictive uncertainty into its epistemic and aleatoric components. First, we show how such a decomposition arises naturally in a Bayesian active learning scenario by following an information theoretic approach. Second, we use a similar decomposition to develop a novel risk sensitive objective for safe reinforcement learning (RL). This objective minimizes the effect of model bias in environments whose stochastic dynamics are described by BNNs with latent variables. Our experiments illustrate the usefulness of the resulting decomposition in active learning and safe RL settings.

Subfields: Artificial Intelligence, Control and Systems Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Control and Systems Engineering

Citations in the corpus

Reliable Uncertainty Estimates in Deep Neural Networks using Noise Contrastive Priors (2018, background): “Priors for neural networks Classic work has investigated entropic priors (Buntine and Weigend, 1991) and hierarchical priors (MacKay, 1992b; Neal, 2012; Lampinen and Vehtari, 2001 2018)), NCPs take the functional view by imposing explicit regularities in terms of the network’s inputs and outputs.”
The Power of Ensembles for Active Learning in Image Classification (2018, background): “Recent literature propose to split the uncertainty over network predictions into a data-dependent aleatoric and a parameter-dependent epistemic uncertainty ([36], [9]).”
Functional Variational Bayesian Neural Networks (2019, background): “[6] Stefan Depeweg, José Miguel Hernández-Lobato, Finale Doshi-Velez, and Steffen Udluft.”
A benchmark environment motivated by industrial control problems (2017, methodology): “An example experiment can be found in [19].”
Noise Contrastive Priors for Functional Uncertainty (2018, background): “More recently, Depeweg et al. (2018) introduce networks with latent variables in order to disentangle forms of uncertainty, and Flam-Shepherd et al. (2017) propose general-purpose weight priors based on approximating Gaussian processes.”
Robust and Efficient Transfer Learning with Hidden Parameter Markov Decision Processes (2017, unknown): “Other approaches use stochastic latent variable inputs to introduce transition stochasticity [12, 30].”
Uncertainty Estimation in Bayesian Neural Networks And Links to Interpretability (2018, background): “In this case, the variational distribution becomes [16]:”
Efficient exploration with Double Uncertain Value Networks (2017, background): “Alltogether, this class of exploration methods usually depends on the ability to learn good transition models (from limited data), a problem which is not trivial itself (Deisenroth and Rasmussen, 2011; Depeweg et al., 2017; Moerland et al., 2017).”