Learning to Condition: A Neural Heuristic for Scalable MPE Inference

Brij Malhotra, Shivvrat Arya, Tahrima Rahman, Vibhav Giridhar Gogate
9/22/2025
cs.LGcs.AI

No official code found. Best implementation: brijml/L2C (0 stars, MIT, last commit 2025-10-23).

Abstract

We introduce learning to condition (L2C), a scalable, data-driven framework for accelerating Most Probable Explanation (MPE) inference in Probabilistic Graphical Models (PGMs), a fundamentally intractable problem. L2C trains a neural network to score variable-value assignments based on their utility for conditioning, given observed evidence. To facilitate supervised learning, we develop a scalable data generation pipeline that extracts training signals from the search traces of existing MPE solvers. The trained network serves as a heuristic that integrates with search algorithms, acting as a conditioning strategy prior to exact inference or as a branching and node selection policy within branch-and-bound solvers. We evaluate L2C on challenging MPE queries involving high-treewidth PGMs. Experiments show that our learned heuristic significantly reduces the search space while maintaining or improving solution quality over state-of-the-art methods.

View on arXivView PDF

Code Implementations(1)

brijml/L2CCommunity56%
MIT

Learning to Condition: A Neural Heuristic for Scalable MPE Inference

00PythonOct 23, 202512 months agoMIT

Similar Papers

Learning to Guide Local Search for MPE Inference in Probabilistic Graphical Models

88% match2026

Learning to Solve the Constrained Most Probable Explanation Task in Probabilistic Graphical Models

83% match2024

EXPLAIN, AGREE, LEARN: Scaling Learning for Neural Probabilistic Logic

81% match2024

Neural Tractability via Structure: Learning-Augmented Algorithms for Graph Combinatorial Optimization

80% match2025

Learning to Think: Information-Theoretic Reinforcement Fine-Tuning for LLMs

80% match1 repo52025

Scalable Coupling of Deep Learning with Logical Reasoning

80% match2023

Cite this paper

@article{malhotra2025learning,
  title  = {Learning to Condition: A Neural Heuristic for Scalable MPE Inference},
  author = {Brij Malhotra and Shivvrat Arya and Tahrima Rahman and Vibhav Giridhar Gogate},
  year   = {2025},
  eprint = {2509.25217},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2509.25217v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2509.25217.svg)](https://codeofpaper.com/paper/2509.25217)

Discussion