In the previous three chapters, we saw how neural networks help directly and indirectly in solving natural language understanding and image processing problems. This is because neural networks are proven to work well with homogeneous data; that is, if all the input features are of the same breed—pixels, words, characters, and so on. On the other hand, when it comes to heterogeneousdata, it is the ensemblemethods that are known to shine. They are well suited to deal with heterogeneous data—for example, where one column contains users' ages, the other has their incomes, and a third has their city of residence.
You can view ensemble estimators as meta-estimators; they are made up of multiple instances of other estimators. The way they combine their underlying estimators is what differentiates between the different ensemble methods—for example, the...
