SVGD as a kernelized Wasserstein gradient flow of the chi-squared divergence.
Sinho ChewiThibaut Le GouicChen LuTyler MaunuPhilippe RigolletPublished in: NeurIPS (2020)
Keyphrases
- chi squared
- gradient flow
- active contours
- information gain
- energy functional
- partial differential equations
- active contour model
- level set
- curve evolution
- variational formulation
- kernel function
- image dependent
- image segmentation
- object boundaries
- level set method
- decision trees
- multiscale
- training data
- energy minimization
- computer vision
- energy function
- text categorization
- feature space
- mutual information