$1\mathrm{.}3\mathrm{.}2$ Error rate versus model complexity

For a fixed training set size, sketch a graph of the typical behavior of training error rate $($y$-$axis$)$ versus model complexity $($x$-$axis$).$ Add to this graph a curve showing the typical behavior of the corresponding test error rate versus model complexity, on the same axes $($again on an IID infinite test set$).$ Show where on the x $-$axis you think is the most complex model that your data supports $($mark this with a vertical line$).$ Choose the x$-$range so that this line is neither on the extreme left nor on the extreme right. Indicate on your vertical axis where zero error is and draw your graphs with increasing error upwards and increasing

complexity rightwards.