Gradients should stay on path: better estimators of the reverse- and forward KL divergence for normalizing flows

Lorenz Vaitl, Kim A. Nicoli, Shinichi Nakajima, Pan Kessel. Gradients should stay on path: better estimators of the reverse- and forward KL divergence for normalizing flows. Mach. Learn. Sci. Technol., 3(4):45006, December 2022. [doi]

@article{VaitlNNK22-0,
  title = {Gradients should stay on path: better estimators of the reverse- and forward KL divergence for normalizing flows},
  author = {Lorenz Vaitl and Kim A. Nicoli and Shinichi Nakajima and Pan Kessel},
  year = {2022},
  month = {December},
  doi = {10.1088/2632-2153/ac9455},
  url = {https://doi.org/10.1088/2632-2153/ac9455},
  researchr = {https://researchr.org/publication/VaitlNNK22-0},
  cites = {0},
  citedby = {0},
  journal = {Mach. Learn. Sci. Technol.},
  volume = {3},
  number = {4},
  pages = {45006},
}