The mean absolute error (MeanAE) and residual sum of squares (RSS) are regression metrics given by the following equations:
The mean absolute error (10.11) is similar to the MSE and MedAE, but it differs in one step of the calculation. The common feature of these metrics is that they ignore the sign of the error and are analogous to variance. MeanAE values are larger than or ideally equal to zero.
The RSS (10.12) is similar to the MSE, except we don't divide by the number of residuals. For this reason, you get larger values with the RSS. However, an ideal fit gives you a zero RSS.
The imports are as follows:
import ch10util import dautil as dl from sklearn import metrics from IPython.display import HTML
Plot the bootstrapped metrics as follows:
sp = dl.plotting.Subplotter(3, 2, context) ch10util.plot_bootstrap('boosting', metrics.mean_absolute_error, sp.ax) sp.label() ch10util.plot_bootstrap...