An approach to regression/density estimation that doesn't require much prior knowledge but only a large amount of data. This includes histograms, kernel smoothing, and nearest-neighbor.
