Question about the probability_combiner param #147
Description
I don't understand what the probability_combiner parameter does in the model.
I was testing the LocalClassifierPerNode and was simply using the default probability_combiner=multiply and it was working well. But when I set probability_combiner=None, the probabilistics coming out of the model started to come very strangely.
I'd expect that by using a classifier by node, there wouldn't be any aggregation, but it doesn't seem to be the case.
Could you clarify what this parameter does ?