Research Papers research paper arxiv statistics machine-learning

Exact and Approximate MCMC for Doubly-intractable Probabilistic Graphical Models Leveraging the Underlying Independence Model

arXivby [Submitted on 4 Oct 2025 (v1), last revised 27 Mar 2026 (this version, v2)]March 30, 20262 min read1 views

arXiv:2510.03587v2 Announce Type: replace-cross Abstract: Bayesian inference for doubly-intractable pairwise exponential graphical models typically involves variations of the exchange algorithm or approximate Markov chain Monte Carlo (MCMC) samplers. However, existing methods for both classes of algorithms require either perfect samplers or sequential samplers for complex models, which are often either not available, or suffer from poor mixing, especially in high dimensions. We develop a method that does not require perfect or sequential sampling, and can be applied to both classes of methods: — Yujie Chen, Antik Chakraborty, Anindya Bhadra

View PDF HTML (experimental)

Abstract:Bayesian inference for doubly-intractable pairwise exponential graphical models typically involves variations of the exchange algorithm or approximate Markov chain Monte Carlo (MCMC) samplers. However, existing methods for both classes of algorithms require either perfect samplers or sequential samplers for complex models, which are often either not available, or suffer from poor mixing, especially in high dimensions. We develop a method that does not require perfect or sequential sampling, and can be applied to both classes of methods: exact and approximate MCMC. The key to our approach is to utilize the tractable independence model underlying the intractable probabilistic graphical model for the purpose of constructing a finite sample unbiased Monte Carlo (and not MCMC) estimate of the Metropolis--Hastings ratio. This innovation turns out to be crucial for scalability in high dimensions. The method is demonstrated on the Ising model. Gradient-based alternatives to construct a proposal, such as Langevin and Hamiltonian Monte Carlo approaches, also arise as a natural corollary to our general procedure, and are demonstrated as well.

Comments: To appear in Proceedings of the 29th International Conference on Artificial Intelligence and Statistics (AISTATS) 2026, Tangier, Morocco

Subjects:

Computation (stat.CO); Methodology (stat.ME); Machine Learning (stat.ML)

Cite as: arXiv:2510.03587 [stat.CO]

(or arXiv:2510.03587v2 [stat.CO] for this version)

https://doi.org/10.48550/arXiv.2510.03587

arXiv-issued DOI via DataCite

Submission history

From: Antik Chakraborty [view email] [v1] Sat, 4 Oct 2025 00:34:25 UTC (272 KB) [v2] Fri, 27 Mar 2026 03:49:11 UTC (480 KB)

Original source

arXiv

https://arxiv.org/abs/2510.03587

Was this article helpful?

Ask AI about this article

Ready

Conversation starters

Ask anything about this article…

Daily AI Digest

Get the top 5 AI stories delivered to your inbox every morning.

More about

researchpaperarxiv

Models

Transformer Paper Authors at AI Startup Debut Open Source Model - Bloomberg.com

Transformer Paper Authors at AI Startup Debut Open Source Model Bloomberg.com

GNews AI transformer

1m4 months ago

Research PapersFresh

RFT FPCM OV - a Hugging Face Space by RFTSystems

huggingface.co RFT FPCM OV - a Hugging Face Space by RFTSystems RFT Fixed Parameter Cosmology Model, Open Validation 1. Fixed‑Parameter Cosmology Panel (FPCM‑OV) This side of the Space shows the core RFT cosmology running on one locked parameter set. Nothing adjusts itself — the whole model stands or falls on this single solution. What people can see here Age at z = 13.67: RFT gives 568.52 Myr , which lines up with JWST early‑galaxy maturity without any tuning. Horizon Ratio: The model naturally produces a horizon about 490× larger than ΛCDM. (This removes the horizon problem without inflation.) Unified Expansion Curve (H_RFT) The purple curve shows how expansion behaves across all redshifts using the same fixed parameters. JWST Maturity Plot The cyan and red lines show how RFT’s predicted