Lifting Architectural Constraints of Injective Flows

• Published in: arXiv.org
• Main Authors: Sorrenson, Peter, Draxler, Felix, Rousselot, Armand, Sander Hummerich, Zimmermann, Lea, Köthe, Ullrich
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 27.06.2024
• Subjects: Free form; Iterative methods; Maximum likelihood estimators; Tables (data); Training

Normalizing Flows explicitly maximize a full-dimensional likelihood on the training data. However, real data is typically only supported on a lower-dimensional manifold leading the model to expend significant compute on modeling noise. Injective Flows fix this by jointly learning a manifold and the...