Frozen-state hierarchical annealing

Area of Science:

• Computational modeling
• Statistical physics
• Image synthesis

Background:

• Synthesizing discrete-state random fields with multi-scale structure is computationally challenging.
• Existing methods struggle to efficiently generate both large- and small-scale features simultaneously.
• Hierarchical methods are being explored to address these limitations.

Purpose of the Study:

• To propose a novel frozen-state approach for hierarchical modeling of discrete-state random fields.
• To improve computational complexity and modeling flexibility in synthesizing multi-scale structures.
• To demonstrate the efficacy of the proposed method in a porous media synthesis problem.

Main Methods:

• A frozen-state hierarchical modeling technique is introduced.
• Simulated annealing is applied at each scale, constrained by parent scale state estimates.
• The domain for annealing at finer scales is restricted to uncertain regions of coarser scales.

Main Results:

• The proposed approach enables the realization of complex structures using simple, local, scale-dependent models.
• Significant improvements in computational complexity are achieved by constraining the annealing domain.
• Successful synthesis of structures in porous media is demonstrated.

Conclusions:

• The frozen-state hierarchical modeling approach offers substantial advantages in flexibility and computational efficiency.
• This method provides an effective solution for synthesizing discrete-state random fields with multi-scale structures.
• The technique is particularly beneficial for complex synthesis problems like those in porous media.