Better parallel - Minor Bug Fix

rsteca / sklearn-deap

Use evolutionary algorithms instead of gridsearch in scikit-learn

MIT License

767 stars 132 forks source link

Better parallel - Minor Bug Fix #45

Closed ryanpeach closed 5 months ago

ryanpeach commented 6 years ago

This prevents FitnessMax and Individual from being created as classes on each worker node, overriding their past declaration.

ryanpeach commented 6 years ago

This all works for me.

flipdazed commented 6 years ago

I found this causes a problem when I iterate a number of instances of EvolutionaryAlgorithmSearchCV.fit(x, y) as in this case I want the previous solution to be re-instantiated as new. Consequently, I had to revert the change after applying.

This is on Windows which may be relevant.