Presently ModelSelection takes a dictionary of "models", which should actually be a dictionary of algorithms.
The functionality to add is to test training a single model/algorithm with multiple datasets.
This resource provides clarity on the difference between an algorithm and a model.
In essence the arguments will likely need to be redefined.
Use cases:
1) User provides multiple algorithms and a single dataset - to select best algorithm
2) User provides single algorithm and multiple datasets - to support feature selection, and test various sample selection scenarios.
3) User provides multiple algorithms and multiple datasets
Presently ModelSelection takes a dictionary of "models", which should actually be a dictionary of algorithms. The functionality to add is to test training a single model/algorithm with multiple datasets.
This resource provides clarity on the difference between an algorithm and a model.
In essence the arguments will likely need to be redefined.
Use cases:
1) User provides multiple algorithms and a single dataset - to select best algorithm 2) User provides single algorithm and multiple datasets - to support feature selection, and test various sample selection scenarios. 3) User provides multiple algorithms and multiple datasets
Objective - support use cases 1 and 2.