How can I specify in a TPOTRegressor or TPOTClassifier to exclude certain algorithms? Let's say, for example, to not generate a pipeline that uses Random Forest or something like that. Is it possible to do this, similarly to what can be done with the TPOT's operators and parameters customization.
yes, you'll want to modify the corresponding config directory to remove the desired modeling technique.
The standard dictionaries can be found here: https://github.com/EpistasisLab/tpot/tree/master/tpot/config
Thanks for the fast response! So basically I can define a tpot_config with all the classifiers and selectors that I want to try and pass it to the TPOTClassifier and than optimize it. Thanks for the clarification
Most helpful comment
yes, you'll want to modify the corresponding config directory to remove the desired modeling technique.
The standard dictionaries can be found here: https://github.com/EpistasisLab/tpot/tree/master/tpot/config