On the Out-of-distribution Generalization of Probabilistic Image ModellingDownload PDF

21 May 2021, 20:45 (modified: 17 Dec 2021, 13:40)NeurIPS 2021 PosterReaders: Everyone
Keywords: OOD detection, OOD generalizationm, Lossless compression, Generative model
TL;DR: We explore the generalization of probabilistic image models in terms of the likelihood and apply the findings to the OOD detecion and lossless comrpession task.
Abstract: Out-of-distribution (OOD) detection and lossless compression constitute two problems that can be solved by the training of probabilistic models on a first dataset with subsequent likelihood evaluation on a second dataset, where data distributions differ. By defining the generalization of probabilistic models in terms of likelihood we show that, in the case of image models, the OOD generalization ability is dominated by local features. This motivates our proposal of a Local Autoregressive model that exclusively models local image features towards improving OOD performance. We apply the proposed model to OOD detection tasks and achieve state-of-the-art unsupervised OOD detection performance without the introduction of additional data. Additionally, we employ our model to build a new lossless image compressor: NeLLoC (Neural Local Lossless Compressor) and report state-of-the-art compression rates and model size.
Supplementary Material: pdf
Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.
15 Replies

Loading