Interventions, Where and How? Experimental Design for Causal Models at Scale
What this work does
Combines posterior uncertainty over graphs and mechanisms with information-based acquisition, selecting both intervention targets and values. Greedy and soft batch selection trade acquisition quality against computation.
Connection to discovery
A baseline for the experiment-action space: choosing what to manipulate and by how much are separate decisions.
Evidence
The consulted derivations and algorithms specify nonlinear additive-noise SCMs, variational inference and bounded intervention-value optimization.
Scope and limitations
Posterior approximation, fixed variables and intervention bounds restrict the search. The review inspected the algorithmic sections rather than independently auditing its empirical comparison.
THE RESEARCH CONVERSATION
Discussion & reactions
Share a question, an insight, or a replication experience. Sign in with GitHub to join the conversation.