# plot training score vs parameter alpha, with x on logscale
# use std over cross-validation folds for error bars
ax = res.plot(x='param_alpha', y='mean_train_score', yerr='std_train_score', logx=True)
# same for test set, plot in the same axes
res.plot(x='param_alpha', y='mean_test_score', yerr='std_test_score', ax=ax)