Built-in Inference Algorithms

Omega comes with a number of built in inference algorithms. You can of course develop your own.

Choosing a Sampling Algorithm

The appropriate sampling algorithm depends on the kind of model.

• If your model is not conditioned, or the conditions are not very restricting, use RejectionSample
• If your model is finite dimensional, continuous and unimodal use NUTS
• If your model is finite dimensional, continuous and multimodal use Replica with NUTS
• If your model is a mixture of discrete and continuous, or not of finite dimension, use SSMH or Replica with SSMH

Conditional Sampling

Conditional sampling is done with rand and the algorithm are selected