How Infer.NET works

Infer.NET works by compiling a model definition into the source code needed to compute a set of inference queries on the model. The diagram below summarises the inference process. 

 
Infer.NET block diagram
 

The steps are:

  1. The user creates a model definition (using the modelling API) and specifies a set of inference queries relating to the model.

  2. The user passes the model definition and inference queries (see Running Inference) to the model compiler, which creates the source code needed to perform those queries on the model, using the specified inference algorithm.  This source code can be written to a file and used directly if required.

  3. The source code is compiled to create a compiled algorithm.  This can be manually executed to get fine-grained control of how inference is performed, or automatically executed via the Infer method.

  4. Using a set of observed values (such as arrays of data), the inference engine executes the compiled algorithm, according to the user-specified settings, so as to produce the marginal distributions requested in the queries.  This can be repeated for different settings of the observed values without recompiling the algorithm.
To learn how to define a model in Infer.NET, go to The Infer.NET modelling API.
©2009-2014 Microsoft Corporation. All rights reserved.  Terms of Use | Trademarks | Privacy Statement