Practical Machine Learning with H2O

Chapter 4. Common Model Parameters

One of the things that makes the H2O APIs so pleasant to use is that each of the machine learning algorithms have much of their interface in common. Later chapters will look at one algorithm at a time, and show how to use them on each of our example data sets. Rather than repeat the same thing in each of those chapters, a lot of the common functionality will be here.

Note

The Python API is object-oriented, which complicates things for this chapter: most of the parameters described here are given when creating the estimator object, but a few are given when calling train() on that object. The latter ones will be pointed out as we go.

The R API (and the underlying REST API) take all parameters in one go.

Each machine learning algorithm will be introduced in its own chapter, but here are their one-line descriptions:

Random Forest

An ensemble (a team) of decision trees. Parameters that apply to it are marked with .

GBM

Gradient Boosting Machines. Another ensemble of decision trees, but with a different approach to random forest. Indicated with .

GLM

Generalized Linear Models. A linear model is the idea of drawing the best straight line through data points. The generalized bit allows it to handle some nonlinearity. Indicated with .

Deep Learning ...