The Art of Machine Learning

7FINDING A GOOD SET OF HYPERPARAMETERS

As discussed in earlier chapters, especially Section 3.2.1, most analysts’ approach to the problem of determining good values of hyperparameters is to use cross-validation. In this chapter, we’ll learn to use a qeML function, qeFT(), that greatly facilitates the process.

7.1 Combinations of Hyperparameters

Note that typically we are talking about sets of hyperparameters. Suppose, for instance, that we wish to use PCA in a k-NN setting. Then we have two hyperparameters: the number of neighbors k and the number of principal components m. Thus we are interested in finding a good combination of a k value and ...