lin-0146

11.2 The BRIDGE learning sketch

For a regime \(e\) with \(P^{(e)}\ll P^{(0)}\), define the Radon–Nikodym ratio

\[ \rho _e(z)=\frac{dP^{(e)}}{dP^{(0)}}(z), \qquad \mathbb E_{P^{(e)}}f = \mathbb E_{P^{(0)}}[\rho _ef]. \]

This identity supplies a statistical interface. It does not identify a causal effect. A differentiable density-ratio or transport model converts the finite regime change into a response field \(v_e\).

The sketch contains:

  • observational and regime-specific probability objects;

  • measure-change or transport arrows from the reference regime;

  • coordinate observations of every response field;

  • the visible distribution \(\mathcal D_V=\operatorname {span}\{ v_1,\ldots ,v_q\} \); and

  • a downstream graph or equivalence-class learner restricted by the geometric mask.

Diagram illustrating 11.2 The BRIDGE learning sketch.