Model ensembling

Combine diverse models through averaging, blending, and stacking to beat any single model. Use when squeezing maximum accuracy from a competition or high-stakes prediction, after strong single models exist.