Generative Modeling of Discrete Joint Distributions by E-Geodesic Flow Matching on Assignment Manifolds.
Bastian BollDaniel Gonzalez-AlvaradoChristoph SchnörrPublished in: CoRR (2024)
Keyphrases
- joint distribution
- matching algorithm
- generative model
- maximum entropy
- marginal distributions
- random variables
- unsupervised learning
- shape matching
- curvature flow
- data sets
- discrete space
- manifold learning
- conditional probabilities
- subband
- linear combination
- higher order
- reinforcement learning
- multiscale
- feature selection
- machine learning