• Clipping is a technique to minimize the influence of extreme outliers.
• In brief, clipping usually caps (reduces) the value of outliers to a specific maximum value. Clipping is a strange idea, and yet, it can be very effective.
• Clipping prevents your model from overindexing on unimportant data. However, some outliers are actually important, so clip values carefully.