{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/adversarial-distillation-of-bayesian-neural","title":"Adversarial Distillation of Bayesian Neural Network Posteriors","arxiv_id":"1806.10317","date":"2018-06-27","proceeding":null,"authors":["Kuan-Chieh Wang","Paul Vicol","James Lucas","Li Gu","Roger Grosse","Richard Zemel"],"abstract":"Bayesian neural networks (BNNs) allow us to reason about uncertainty in a\nprincipled way. Stochastic Gradient Langevin Dynamics (SGLD) enables efficient\nBNN learning by drawing samples from the BNN posterior using mini-batches.\nHowever, SGLD and its extensions require storage of many copies of the model\nparameters, a potentially prohibitive cost, especially for large neural\nnetworks. We propose a framework, Adversarial Posterior Distillation, to\ndistill the SGLD samples using a Generative Adversarial Network (GAN). At\ntest-time, samples are generated by the GAN. We show that this distillation\nframework incurs no loss in performance on recent BNN applications including\nanomaly detection, active learning, and defense against adversarial attacks. By\nconstruction, our framework not only distills the Bayesian predictive\ndistribution, but the posterior itself. This allows one to compute quantities\nsuch as the approximate model variance, which is useful in downstream tasks. To\nour knowledge, these are the first results applying MCMC-based BNNs to the\naforementioned downstream applications.","url_abs":"http://arxiv.org/abs/1806.10317v1","url_pdf":"http://arxiv.org/pdf/1806.10317v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"adversarial-distillation-of-bayesian-neural","repo_url":"https://github.com/wangkua1/apd_public","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"active-learning","task_name":"Active Learning"},{"task_slug":"anomaly-detection","task_name":"Anomaly Detection"},{"task_slug":null,"task_name":"Generative Adversarial Network"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1806.10317","atlas_url":"https://app.syntology.ai/?focus=1806.10317","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}