Form implementation

Strategy Choice

First, you need to add a tuple in the form (acronym, full name) to the alambic_app > constantes.py script in the list of the strategy choice AL_ALGORITHMS_CHOICES, as well in the form of a acronym: full name entry in the dictionary AL_ALGORITHMS_MATCH.

This operation will allow your strategy to be chosen during the setup phase, as well as showing the full name in the plots.

Parametrization

Currently, the query strategies do not have parametrization available. Thus, no additional form are necessary.

Strategy implementation

Preface

Our framework is currently based on the library ALiPy (repository here).

Additional strategy must be built with the BaseQueryIndex class from that very same library and implement the select method.

Location

Implemented strategies must be implemented in the alambic_app > active_learning > strategies.py file.

Once implemented, the model must be imported and added to the AL_ALGORITHMS_MATCH dictionary in the alambic_app > machine_learning > setup.py script with their corresponding acronym (see Strategy Choice in the previous section).

Parametrization

If specific parameters must be added to the strategy, you can add the code in the alambic_app > machine_learning > setup.py script, in the set_query_strategy method.

Documentation

Update the documentation ! It is present in the folder docs in the query_strategy.md document.

Checklist

• Add the (acronym, full name) tuple in the AL_ALGORITHMS_CHOICES list and the dictionary AL_ALGORITHMS_MATCH in the constantes.py script
• (Opt.) Implement the strategy if not existing in the alambic_app > active_learning > strategies.py file.
• Import in alambic_app > machine_learning > setup.py and add to AL_ALGORITHMS_MATCH with the acronym as key and the model function as value
• Modify the set_query_strategy method if needed for the parameters of the strategy.
• Update the documentation (it’s important too !)
• Be proud of yourself, you did it ! (Well, after tons of debug and testing, of course)

---