Bayesian Variational Autoencoders for Unsupervised Out-of-Distribution Detection
By Erik A. Daxberger, José Miguel Hernández-Lobato. 2019
69 citations recorded by the source corpus.
Despite their successes, deep neural networks may make unreliable predictions when faced with test data drawn from a distribution different to that of the training data, constituting a major problem for AI safety. While this has recently motivated the development of methods to detect such out-of-distribution (OoD) inputs, a robust solution is still lacking. We propose a new probabilistic, unsupervised approach to this problem based on a Bayesian variational autoencoder model, which estimates a full posterior distribution over the decoder parameters using stochastic gradient Markov chain Monte Carlo, instead of fitting a point estimate. We describe how information-theoretic measures based on this posterior can then be used to detect OoD inputs both in input space and in the model's latent space. We empirically demonstrate the effectiveness of our proposed approach.
Subfields: Artificial Intelligence
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
- Adrian Weller
- Umang Bhatt
- Austin Tripp
- Erik A. Daxberger
- Javier Antorán
- Tameem Adel
- José Miguel Hernández-Lobato
- Dana Rezazadegan
Citations in the corpus
- Getting a CLUE: A Method for Explaining Uncertainty Estimates (2020, methodology): “It enables the former to be used for uncertainty aware tasks, such as OOD detection (Daxberger & Hernández-Lobato, 2019), continual learning (Nguyen et al., 2018), active learning (Depeweg et al., 2018), and Bayesian optimization (Springenberg et al., 2016).”
- A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challenges (2020, background): “[135] tried to detect OoD samples during test phase.”
- Sample-Efficient Optimization in the Latent Space of Deep Generative Models via Weighted Retraining (2020, methodology): “LSO using DGMs was first applied to chemical design [20], and further built upon subsequently [28, 37, 15, 31, 9, 10, 22, 43].”