The experiment estimator::cpp can be run either on command line, or as a mex file in Matlab.
In order to use it for estimation of an ARX model, we can define the following
User Infos and their use structure:
The structure is interpreted by application estimator
, which looks for fields:
- system
- description of a Data Source generating Data. The structure must by UI with
type="DS_offspring"
. In our example, it is of type "ArxDS" which is parsed by bdm::UIArxDS UIbuilder generating a Data Source simulating ARX process.
- estimator
- description of a Baysian model used to estimate parameters of the data model. In this case, it is of type "ARXest" which is parsed by bdm::UIARX UIbuilder generating Bayesian estimator of autoregressive processess.
- logger
- description of a way how to store results. UI is of
type="logger_offspring"
. In this case, it is of class "dirfilelog" which is parsed by bdm::UIdirfilelog which generates object storing data in directory specified by dirname="" field in fileformat understood by program kst.
When the application estimator is run with the above code, it produces a directory of data files, which can be displayed by program kst. Expected results are:
The matlab mex file can be run with exactly the same configuration as above. However, when we wish to see the results in Matlab, we may wish to change the logger object to
type="mexlog"
which will store the results in a matlab structure.
The exact configuration file may look as follows:
The resulting structure can be displayed using matlab script arx_test_disp.m, typically producing the following results: