Locally Most Powerful Bayesian Test for Out-of-Distribution Detection using Deep Generative Models
Dec 6, 2021
Speakers
Keunseo Kim
Speaker · 0 followers
Junchul Shin
Speaker · 0 followers
Heeyoung Kim
Speaker · 0 followers
About
Several out-of-distribution (OOD) detection scores have been recently proposed for deep generative models, as the direct use of the likelihood threshold for OOD detection has shown to be problematic. In this paper, we propose a new OOD score based on a Bayesian hypothesis test called the locally most powerful Bayesian test (LMPBT). The LMPBT is locally most powerful in the sense that the alternative hypothesis (representative parameter for OOD sample) is specified so as to maximize the probability that the Bayes factor exceeds the evidence threshold in favor of the alternative hypothesis as long as the parameter specified under the alternative hypothesis is in a neighborhood of the parameter specified under the null hypothesis. That is, under this neighborhood parameter condition, the test with the proposed alternative hypothesis maximizes the probability of correct detection of OOD samples. We also propose numerical strategies for more efficient and reliable computation of the LMPBT for practical application to deep generative models. We evaluate the OOD detection performance of the LMPBT on various benchmark datasets and demonstrate its superior performance over existing OOD detection methods.Several out-of-distribution (OOD) detection scores have been recently proposed for deep generative models, as the direct use of the likelihood threshold for OOD detection has shown to be problematic. In this paper, we propose a new OOD score based on a Bayesian hypothesis test called the locally most powerful Bayesian test (LMPBT). The LMPBT is locally most powerful in the sense that the alternative hypothesis (representative parameter for OOD sample) is specified so as to maximize the probabili…
Organizer
NeurIPS 2021
Account · 1.9k followers
About NeurIPS 2021
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Dealing With Misspecification In Fixed-Confidence Linear Top-m Identification
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias
Watch later
Yufei Xu, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
The Collective Intelligence of Army Ants, and the Robots They Inspire
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Q&A 4
Watch later
Felipe Tobar, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Performance-optimized neural networks as an explanatory framework for decision confidence
Watch later
Hakwan Lau, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Faster Directional Convergence of Linear Neural Networks under Spherically Symmetric Data
Watch later
Dachao Lin, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%