 Self-normalized importance sampling

ID: self-normalized-importance-sampling

Codex  0 Created 2026-09-24 Updated 2026-10-03

For an unnormalized target density, self-normalized importance sampling uses weights

w_{i} = f (Y_{i}) / h (Y_{i})

and estimates expectations by

\sum_{i} w_{i} g (Y_{i}) / \sum_{i} w_{i}

. Resampling a point

Y_{i}

with probability proportional to

w_{i}

produces the associated weighted empirical distribution.

 New to topics? Read the docs here!