From Diffusion to Drifting: Generative Modeling as Learned Distributional Transport
Published:
A note on drifting models, Wasserstein gradient flows, and why one-step generators can be viewed as training-time distributional transport.
Posts by Tags
diffusion
estimation
Why Any Distribution-Free Lower-Bound Estimator for Mutual Information Can’t Beat log N
Published:
A short explanation of McAllester & Stratos’ (2020) mututal information estimation result plus some intuition.
generative-models
From Diffusion to Drifting: Generative Modeling as Learned Distributional Transport
Published:
A note on drifting models, Wasserstein gradient flows, and why one-step generators can be viewed as training-time distributional transport.
impossibility
Why Any Distribution-Free Lower-Bound Estimator for Mutual Information Can’t Beat log N
Published:
A short explanation of McAllester & Stratos’ (2020) mututal information estimation result plus some intuition.
mutual-information
Why Any Distribution-Free Lower-Bound Estimator for Mutual Information Can’t Beat log N
Published:
A short explanation of McAllester & Stratos’ (2020) mututal information estimation result plus some intuition.
optimal-transport
From Diffusion to Drifting: Generative Modeling as Learned Distributional Transport
Published:
A note on drifting models, Wasserstein gradient flows, and why one-step generators can be viewed as training-time distributional transport.
sinkhorn
From Diffusion to Drifting: Generative Modeling as Learned Distributional Transport
Published:
A note on drifting models, Wasserstein gradient flows, and why one-step generators can be viewed as training-time distributional transport.