On Last-Layer Algorithms for Classification: Decoupling Representation from Uncertainty Estimation
By Nicolas Brosse, Carlos Riquelme, Alice Martin, Sylvain Gelly, Eric Moulines. 2020
37 citations recorded by the source corpus.
Uncertainty quantification for deep learning is a challenging open problem. Bayesian statistics offer a mathematically grounded framework to reason about uncertainties; however, approximate posteriors for modern neural networks still require prohibitive computational costs. We propose a family of algorithms which split the classification task into two stages: representation learning and uncertainty estimation. We compare four specific instances, where uncertainty estimation is performed via either an ensemble of Stochastic Gradient Descent or Stochastic Gradient Langevin Dynamics snapshots, an ensemble of bootstrapped logistic regressions, or via a number of Monte Carlo Dropout passes. We evaluate their performance in terms of \emph{selective} classification (risk-coverage), and their ability to detect out-of-distribution samples. Our experiments suggest there is limited value in adding multiple uncertainty layers to deep classifiers, and we observe that these simple methods strongly outperform a vanilla point-estimate SGD in some complex benchmarks like ImageNet.
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
- Agustinus Kristiadi
- Philipp Hennig
- Farid Boussaïd
- Mohammed Bennamoun
- Matthias Hein
- Hamid Laga
- Laurent Valentin Jospin
- Wray L. Buntine
In Control and Systems Engineering
Citations in the corpus
- Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks (2020, methodology): “Our finding agrees with the prior observation that last-layer Bayesian approximations are often sufficient (Brosse et al., 2020).”
- Hands-On Bayesian Neural Networks—A Tutorial for Deep Learning Users (2020, background): “Instead, recent research tried to use only a few stochastic layers, usually positioned at the end of the networks [8, 97].”