normalizing flows
Normalizing flows are a class of generative models that enable complex distributions to be represented as transformations of simpler distributions. In AI, they allow for efficient sampling and likelihood estimation in probabilistic modeling contexts.
- Asymptotically exact variational flows via involutive MCMC kernels
- CDFlow: Building Invertible Layers with Circulant and Diagonal Matrices
- Detecting Generated Images by Fitting Natural Image Distributions
- Flexible Language Modeling in Continuous Space with Transformer-based Autoregressive Flows
- Multivariate Latent Recalibration for Conditional Normalizing Flows
- Normalizing Flows are Capable Models for Continuous Control
- Path Gradients after Flow Matching
- STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesis