[ldd-ml] Address suggestions from Ed Guinness

[wkiri]

Issue Type Enhancement

Describe the issue identified @edguinness provided these suggestions for the ML LDD:

General comment The dictionary seems to be focused on image applications based on the attributes and classes. Maybe that could be emphasized in a comment field at the beginning of the dictionary ingest file and any other documentation that goes with the dictionary.

Attributes

1. Should data_set_size have data type of ASCII_Integer because it is a number of images? It does not seem that is would be a real number.
2. Are there formation rules for data_set_version_id? For example version_id in the PDS4 common dictionary has the form of M.n, where the M is the major version and n is used for minor updates. See section x.x of the PDS4 standards reference for details. Same comment for trained_model_version_id.
3. Only one enumerated value is defined for algorithm_type while the description lists several. Are more values expected to be added later?

Classes Both the Trained_Machine_Learning_Model_Class and Machine_Learning_Algorithm classes have element_flag set to true. A PDS4 dictionary best practice is to only have one class with this flag set to true. That means there should be a top-level enclosing class that includes all the lower-level classes. Describe the issue identified (if applicable) A clear and concise description of the issue you encountered using the dictionary.

LDD Dictionary Version 1.0.0.0

PDS4 IM Version 1G00

[wkiri]

I created an enclosing class called Machine_Learning with element_flag set to true. All other classes now have element_flag set to false. This update pre-dates when this issue was opened. See:

• https://github.com/pds-data-dictionaries/ldd-ml/commit/e7b2329c85b42f69b48577e0970276efdca42ce1
• https://github.com/pds-data-dictionaries/ldd-ml/commit/afcb7b78cd94f1848757f81b780e2b2ea6b8ffca