Skip to content

Active Intervention Selection Strategies

Home / Open Questions / Active Intervention Selection Strategies

Background: Causal discovery methods rely on intervention sets to distinguish between causal pathways, with efficiency and informativeness being key factors in computational cost. Simulator-based approaches in particular are bounded by the high computational demand of evaluating complex physical systems.

Question / Future Work: There is an unresolved need for active intervention selection strategies that optimize which variables to clamp or manipulate in a simulator to minimize computational costs while maximizing the distinguishability of causal models. This involves developing adaptive target selection that accounts for both the estimated graph structure and the simulator’s computational budget.

Why It Matters: Key for scaling causal discovery to complex scientific systems where simulation budgets are a primary constraint.

Metadata & Links