I think it is cool to add that the cooler name for stacking is deep learning.

Neural networks are just an easier (and more homogeneous) way of implementing stacking. The network is just a stack of weak-classifier arrays (neurons). The name multi-layer perceptron (MLP) is just that.

However, beware that Kaggle is not a good proxy of the real world. Voting and stacking are great tools to get that extra 1% to win a competition, but seldom are good choices in practice. The added complexity is rarely worth it.