DAG-DGPs place GP priors on functions along a DAG, prove non-collapse bounds under separating nodes and intermediate observations, and give structured VI that retains compositional uncertainty and explaining-away, with SOTA multi-fidelity results.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2026 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Deep Gaussian Processes on Directed Acyclic Graphs
DAG-DGPs place GP priors on functions along a DAG, prove non-collapse bounds under separating nodes and intermediate observations, and give structured VI that retains compositional uncertainty and explaining-away, with SOTA multi-fidelity results.