scikit-learn Cookbook , Second Edition - Second Edition

There's more...

Underneath it all, the kernel computes covariances between points in X. It assumes that similar points in the inputs should lead to similar outputs. Gaussian processes are great for confidence predictions and smooth-like outputs. (Later, we will see random forests, that do not lead to smooth outputs even though they are very predictive.)

We may need to understand the uncertainty in our estimates. If we pass the eval_MSE argument as true, we'll get MSE and the predicted values, so we can make the predictions. A tuple of predictions and MSE is returned, from a mechanics standpoint:

test_preds, MSE = gpr7n.predict(boston_X[~train_set], return_std=True)MSE[:5]array([ 1.20337425, 1.43876578, 1.19910262, 1.35212445, 1.32769539]) ...