Pattern completion is the ability of a neural network to reconstruct correct output from incomplete input. For example, we can see a picture of a dog with its whole body obscured. The tail is all we need to conclude "this is a dog!". Similarly, the dog may show up in foggy darkness and still be recognized. You can recognize the dog even if it dresses up like your mother-in-law.