Practical Modelling of Mixed-Tailed Data with Normalizing Flows

Open Access
Authors
Publication date 2024
Journal Transactions on Machine Learning Research
Article number 2998
Volume | Issue number 2024 | 10
Number of pages 19
Organisations
  • Faculty of Economics and Business (FEB) - Amsterdam Business School Research Institute (ABS-RI)
  • Faculty of Science (FNWI) - Informatics Institute (IVI)
Abstract Capturing the correct tail behavior is difficult, yet essential for a faithful generative model. In this work, we provide an improved framework for training flows-based models with robust capabilities to capture the tail behavior of mixed-tail data. We propose a combination of a tail-flexible base distribution and a robust training algorithm to enable the flow to model heterogeneous tail behavior in the target distribution. We support our claim with extensive experiments on synthetic and real world data.
Document type Article
Language English
Published at
https://openreview.net/forum?id=uphsKDj0Uu (Final published version)
Other links
Downloads
Permalink to this page
Back