Why are there multiple hypotheses in ensemble methods

ensemble_methods

#1

Hello,

While reading about ensemble methods I came upon the following paragraph:

I have a few questions regarding the highlighted parts,which are as follows:
1.What is this space of Hypotheses?
2.Does algorithms really generate many different H’s?
3.Again there is a true hypothesis,so the others were not true but still used by the method,how?

My confusion is from my understanding that there is generally one hypothesis,so how come there are multiple here.Or is it the case that since independent variables are many,they can give different values of Y which are taken to be as the Hypotheses??
Can somebody please clarify these points!!


#2

@Pagal_guy,

As we know that we develop a model to know the unknown class of test data set and model is developed using training data set. Here prediction is based on single model.

In Ensemble modeling, we apply crowd wisdom to perform any classification task i.e. we use collection of possible models to predict the outcome. This collection of model can regard as Hypothesis Space and each single point in this hypothesis space corresponds to a single model. A classification algorithm generally makes certain assumptions about this hypothesis space to select right model. The algorithm also defines the rules to measure the quality of model so that the model has the nest measure in the hypothesis space will be returned.

For more detail, please refer this link:

Regards,
Steve