Nonparametric Involutive Markov Chain Monte Carlo
summary
Framework generalizes involutive MCMC to universal PPLs and implements a concrete fix for trans‑dimensional moves: it appends samples until proposal has positive density before acceptance, adjusts the MH ratio accordingly, and requires an auxiliary kernel per dimension plus involutions satisfying the A3 involution property. Eight of 15 standard involutive samplers lift under these assumptions. Preliminary experiments on an infinite Gaussian‑mixture (30 points from 3 Gaussians) show the NP extension recovers the true k more reliably than baselines.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:26Metropolis Hasting
- 6:04Nonparametric MH
- 11:52Assumptions
- 13:35Extensions
- 15:26Experiments
- 16:36Summary
- 17:23Conclusion
- 18:22Question
related talks
12mIterative Methods and High-Dimensional Statistics
51mRandom Projections for Probabilistic Inference
14mSequential Monte Carlo Program Synthesis with Refinement Proposals
01h 06mNonparametric Bayesian Methods: Models, Algorithms, and Applications I
01h 04mNonparametric Bayesian Methods: Models, Algorithms, and Applications II
19mIncremental Computation for Efficient Programmable Inference in Probabilistic Programs
29mMotivating Probabilistic Programming
57mA New Perspective on High-Dimensional Causal Inference