Grid Search for Parameter Optimisation

[lm-fsng]

An option for VariantSpark to search for optimal parameters using a grid search where users can provide a search grid with parameters to test. There is an implementation in Python in the scikit-learn python library (https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html).

This will help with the parameter optimisation process.

[rocreguant]

Two main points:

1. How to select the best model?
2. Automated way to find the best VS model

Current options:

• Piotr's ranger OOB.
• Other options, are against the benchmark and true positives.
• LocalFDR (Rob's method)
• Rule of thumb (the results seem quite consistent)