Energetic variational inference derives existing particle-based VI methods from energy-dissipation laws and proposes an approximation-then-variation scheme that preserves particle-level structure while reducing KL divergence.
In: International Conference on Machine Learning, pp
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2020 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Particle-based Energetic Variational Inference
Energetic variational inference derives existing particle-based VI methods from energy-dissipation laws and proposes an approximation-then-variation scheme that preserves particle-level structure while reducing KL divergence.