A Unifying View of Variational Generative Wasserstein Flows
Two new research papers explore advanced techniques in generative modeling. The first paper introduces Generative Wasserstein Flows (GWF) as a unified framework for various generative models, extending to new algorithms and clarifying connections with GANs. The second paper proposes using Koopman operators to linearize continuous normalizing flows, enabling faster sampling and new analytical insights into the generative process. AI
IMPACT These papers introduce novel theoretical frameworks and methods that could advance generative modeling capabilities and efficiency.