Awesome Normalizing Flows

Overview

Awesome Normalizing Flows provides comprehensive resources on normalizing flows, a class of generative models with exact likelihood computation.

What are Normalizing Flows?

Normalizing flows are a family of generative models that learn an invertible transformation from a simple distribution to a complex distribution, enabling:

• Exact likelihood computation
• Efficient sampling
• Bijective transformations

Key Architectures

• Autoregressive Flows - MAF, IAF
• Coupling Flows - RealNVP, Glow
• Residual Flows - Residual normalizing flows
• Continuous Flows - Neural ODEs, FFJORD
• Graph Flows - For structured data

Applications

• Density Estimation - Modeling complex probability distributions
• Generative Modeling - Image and audio generation
• Variational Inference - Posterior approximation
• Anomaly Detection - Out-of-distribution detection
• Molecular Modeling - Drug discovery and materials science

Research Papers

• Foundational papers on flow-based models
• Recent advances in architecture design
• Application-specific innovations
• Theoretical analysis

Implementations

• PyTorch implementations
• TensorFlow implementations
• JAX-based frameworks
• Pre-trained models

Use Cases

• Research in generative modeling
• Probabilistic machine learning
• Scientific computing
• Computer vision applications
• Molecular design

Pricing

Free and open-source resource collection.