CPE is a normalizing-flow posterior estimator whose architecture is sparsified according to the conditional dependencies of the model's prior and posterior graphs, with a rectified-flow objective enabling 20-step Euler sampling.
Its generative process is defined as: θ∼U 2(−10,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Causal Posterior Estimation
CPE is a normalizing-flow posterior estimator whose architecture is sparsified according to the conditional dependencies of the model's prior and posterior graphs, with a rectified-flow objective enabling 20-step Euler sampling.