Subclasses of Class Function used to Implement Transformations of Statistical Models
This work is incremental, focusing on software engineering improvements for a specific MML-based machine learning library.
The paper tackles the implementation of transformations for statistical models within an existing Minimum Message Length (MML) inference library, defining subclasses of class Function to achieve desirable object-oriented and mathematical properties.
A library of software for inductive inference guided by the Minimum Message Length (MML) principle was created previously. It contains various (object-oriented-) classes and subclasses of statistical Model and can be used to infer Models from given data sets in machine learning problems. Here transformations of statistical Models are considered and implemented within the library so as to have desirable properties from the object-oriented programming and mathematical points of view. The subclasses of class Function needed to do such transformations are defined.