Samplers

Samplers drive data generation by exploring configuration space with QBC uncertainty quantification.

Available sampler modules

ALF currently includes several sampler-related modules:

1. alframework.samplers.mlmd_sampling Main uncertainty-driven sampling workflow using Langevin dynamics. This is the primary sampler used in many ALF runs.

2. alframework.samplers.reactive_sampler Reactive sampling workflows based on NEB and dimer-style methods for constructing reaction-focused datasets.

Choosing a sampler

Use mlmd_sampling for general active-learning loops where uncertainty thresholds are used to decide when to call QM.

Use reactive_sampler when the target problem is reaction-pathway sampling and transition-state exploration.

Sampler API links

You can link from this guide directly to API pages:

• Samplers package API

• MLMD sampling module

• Reactive sampler module

Related examples

See Simple Water Example and Molten Salt Example for MLMD-style sampling, and Reactive Sampling Example for reaction-path sampling.