Adversarial Distillation of Bayesian Neural Network Posteriors

By Kuan-Chieh Jackson Wang, Paul Vicol, James Lucas, Li Gu, Roger Baker Grosse, Richard S. Zemel. 2018

61 citations recorded by the source corpus.

Bayesian neural networks (BNNs) allow us to reason about uncertainty in a principled way. Stochastic Gradient Langevin Dynamics (SGLD) enables efficient BNN learning by drawing samples from the BNN posterior using mini-batches. However, SGLD and its extensions require storage of many copies of the model parameters, a potentially prohibitive cost, especially for large neural networks. We propose a framework, Adversarial Posterior Distillation, to distill the SGLD samples using a Generative Adversarial Network (GAN). At test-time, samples are generated by the GAN. We show that this distillation framework incurs no loss in performance on recent BNN applications including anomaly detection, active learning, and defense against adversarial attacks. By construction, our framework not only distills the Bayesian predictive distribution, but the posterior itself. This allows one to compute quantities such as the approximate model variance, which is useful in downstream tasks. To our knowledge, these are the first results applying MCMC-based BNNs to the aforementioned downstream applications.

Subfields: Efficiency & Compression, Probabilistic & Bayesian Methods

How researchers see this paper

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

In Efficiency & Compression

In Probabilistic & Bayesian Methods

Citations in the corpus

Laplace Redux - Effortless Bayesian Deep Learning (2021, methodology): “Moreover, while unbiased compared to its simpler alternatives, MCMC methods are notoriously expensive in practice and, thus, often require further approximations such as distillation [82, 83].”
Approximating the Predictive Distribution via Adversarially-Trained Hypernetworks (2018, methodology): “This is similar to what is proposed by Wang et al. [34], but the scalability of such an approach might be limited as the dimensionality of W rapidly increases and a huge amount of samples from p ( W | D ) would be required to capture the actual shape of this distribution with the hypernetwork .”
Hands-On Bayesian Neural Networks—A Tutorial for Deep Learning Users (2020, methodology): “A Bayesian teacher can also be used to compress a large set of samples generated using MCMC [111].”
A survey of uncertainty in deep neural networks (2021, unknown): “[180] showed that Generative Adversarial Models”
HyperGAN: A Generative Model for Diverse, Performant Neural Networks (2019, methodology): “To do this, a standard approach would be to acquire a large set of trained model parameters and use those as training data to the GAN (Wang et al., 2018).”
Pitfalls of In-Domain Uncertainty Estimation and Ensembling in Deep Learning (2020, background): “Some works suggest construction of ensembles of stochastic computation graphs (Tomczak et al., 2018), while others make use of the collected snapshots to construct a stochastic computation graph (Wang et al., 2018; Maddox et al., 2019).”