Hands-On Ensemble Learning with Python

Learning curves

Another way to identify bias and variance is to generate learning curves. Like validation curves, we generate a number of in-sample and out-of-sample performance statistics with cross-validation. Instead of experimenting with different hyperparameter values, we utilize different amounts of training data. Again, by examining the means and standard deviations of in-sample and out-of-sample performance, we can get an idea about the amount of bias and variance inherent in our models.

Scikit-learn implements learning curves in the sklearn.model_selection module as learning_curve. Once again, we will use the KNeighborsClassifier example from Chapter 1, A Machine Learning Refresher. First, we import the required libraries and load ...