Get Started
Research questionHow should experiments be selected when nuisance uncertainty makes structural explanations only partially identifiable?An observation can appear highly informative while leaving multiple structural explanations indistinguishable across nuisance conditions. Selecting experiments by information gain alone may therefore produce poor structural resolution or unjustified exclusions.
Machine Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Resolution-Aware Experimental Design under Partial IdentifiabilityEvidence comes from a learned score-based implementation evaluated under constrained sensing on fluvial and mechanistic methane-oxidation subsurface-flow benchmarks. The reported results include disagreements with expected-information-gain selection and a finite-sample, tail-sensitive nuisance-risk guarantee at a prospectively specified 5% false-exclusion tolerance in the methane-oxidation benchmark; broader settings are not established.research paper · Sep 3, 2026
Related questions
How can differentiable causal discovery override wrong edge priors while preserving directional identifiability?How should sequential hypothesis tests choose sensing actions with variable costs under an error constraint?How can we reliably select a heterogeneous treatment effect estimator without observing treatment effects?How should we choose graph counterfactual explainers when small edits, realism, coverage, and quality conflict?
Home
Topics
Search
Library