New algorithms for calculating partition functions on random graphs
Researchers found zero-free regions for specific models on random regular graphs.
Researchers studied antiferromagnetic spin systems on random regular graphs. They looked at the hardcore model and the antiferromagnetic Potts model.
The study found zero-free regions throughout the interior of the replica-symmetric regime. This applies to both models on random -regular graphs.
For the hardcore model, the regime extends up to tree occupation density . For -colorings, it extends up to degree .
The work uses reinforced small subgraph conditioning. This method resolves short-cycle fluctuations in the complex plane by lifting classical small subgraph conditioning to random holomorphic functions.
They also overcame the standard second-moment barrier. This barrier is caused by the exponential blowup of the bare second moment inside the replica-symmetric phase.
The researchers obtained deterministic polynomial-time algorithms. These algorithms multiplicatively approximate the corresponding partition functions at fixed rational parameters.
Key facts
- Zero-free regions were established for the hardcore model and the antiferromagnetic Potts model.
- New algorithms provide multiplicative approximations of partition functions.
- The results extend beyond the parameter ranges reached by previously known search algorithms.