Running inference

This section explains how to choose a pre-built sampler, configure a run, and interpret its results and evidence estimate (normalizing constant). Custom pipelines can also be assembled from composable steps.

Configuring inference

Start with the available pre-built samplers, then configure the SMC runner. For sequential data or particle filtering, see Using observations for passing an observation stream to the runner.

Contents

• Pre-built samplers
• Configuring an SMC runner
• Built-in callbacks
• Using observations

Understanding outputs

After a run, inspect the particle population, extract posterior samples, and check diagnostics. The evidence guide explains the normalizing constant and the assumptions behind its uncertainty estimates.

Contents

• Working with results
• Evidence and uncertainty