Timezone: »

 
Poster
Why Normalizing Flows Fail to Detect Out-of-Distribution Data
Polina Kirichenko · Pavel Izmailov · Andrew Wilson

Tue Dec 08 09:00 AM -- 11:00 AM (PST) @ Poster Session 1 #331

Detecting out-of-distribution (OOD) data is crucial for robust machine learning systems. Normalizing flows are flexible deep generative models that often surprisingly fail to distinguish between in- and out-of-distribution data: a flow trained on pictures of clothing assigns higher likelihood to handwritten digits. We investigate why normalizing flows perform poorly for OOD detection. We demonstrate that flows learn local pixel correlations and generic image-to-latent-space transformations which are not specific to the target image datasets, focusing on flows based on coupling layers. We show that by modifying the architecture of flow coupling layers we can bias the flow towards learning the semantic structure of the target data, improving OOD detection. Our investigation reveals that properties that enable flows to generate high-fidelity images can have a detrimental effect on OOD detection.

Author Information

Polina Kirichenko (New York University)
Pavel Izmailov (New York University)
Andrew Wilson (New York University)

I am a professor of machine learning at New York University.

More from the Same Authors