Embedded-model flows: Combining the inductive biases of model-free deep learning and explicit probabilistic modeling
arXiv:2110.06021
Abstract
Normalizing flows have shown great success as general-purpose density estimators. However, many real world applications require the use of domain-specific knowledge, which normalizing flows cannot readily incorporate. We propose embedded-model flows (EMF), which alternate general-purpose transformations with structured layers that embed domain-specific inductive biases. These layers are automatically constructed by converting user-specified differentiable probabilistic models into equivalent bijective transformations. We also introduce gated structured layers, which allow bypassing the parts of the models that fail to capture the statistics of the data. We demonstrate that EMFs can be used to induce desirable properties such as multimodality, hierarchical coupling and continuity. Furthermore, we show that EMFs enable a high performance form of variational inference where the structure of the prior model is embedded in the variational architecture. In our experiments, we show that this approach outperforms state-of-the-art methods in common structured inference problems.
References in corpus (10)
- NICE: Non-linear Independent Components Estimation
- Edward: A library for probabilistic modeling, inference, and criticism
- Estimation of Thermodynamic Observables in Lattice Field Theories with Deep Generative Models
- Deep Probabilistic Programming
- Equivariant Hamiltonian Flows
- Introduction to Normalizing Flows for Lattice Field Theory
- Targeted Likelihood-Free Inference of Dark Matter Substructure in Strongly-Lensed Galaxies
- Joint Distributions for TensorFlow Probability
- Dark Matter Subhalos, Strong Lensing and Machine Learning
- Normalizing Flows as a Novel PDF Turbulence Model