Since running a machine-learning algorithm is a CPU-intensive task for all but trivial datasets, the algorithms themselves are implemented in C. In order to execute the algorithms from the python environment, wrapper are provided. A wrapper is a subclass of the ML_Algorithm class designed to run an particular binary application and provide a consistent interface to the python side.
To create a wrapper, one need only construct a new instance of the appropriate wrapper type. All wrapper have the same interface for their constructors: <name>(Dataset, parameters) where Dataset is any Dataset or View object and parameters is a python hash which contains the parameters as key/value pairs.
Each wrapper has unique requirements in terms of required parameters and environment variable. You should refer to the API documentation for full details, but the essentials will be summarized at the top of each wrapper section. Please, note that the summary reflected the absolute minimum requirements, there may be many optional parameters, or certain values of the required parameters may require other parameters to be set.