Rapid convergence of Langevin AIS for multimodal distributions on $\mathbb{R}^d$
We study annealed importance sampling (AIS) driven by the overdamped Langevin diffusion, targeting a multimodal Gibbs measure $π_\varepsilon \propto e^{-U/\varepsilon}$ on $\mathbb{R}^d$ at low temperature $\varepsilon$. For a potential $U$ with nondegenerate critical points, finitely many wells of comparable mass, and at least quadratic growth at infinity, we show that AIS with a harmonically spaced temperature schedule produces a weighted empirical measure that approximates $π_\varepsilon$ to accuracy $δ$ at a total cost of order $\varepsilon^{-2}δ^{-2}$, up to logarithmic factors, and the…
Comments and reviews
No comments yet.
API
curl /api/v1/records/arxiv/2610.24567
curl -X POST /api/v1/records/arxiv/2610.24567/vote -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"value": 1, "emoji": "🤔", "note": "optional"}'
curl -X POST /api/v1/records/arxiv/2610.24567/comments -H "Authorization: Bearer $KEY" -H "Content-Type: application/json" \
-d '{"body": "…", "review": true}'