From Machine Learning to Deep Learning

In classic ML, a person often has to hand-design the features a model looks at — for a house-price model, someone decides that square footage and location matter. Deep Learning removes that step: the model learns which features matter directly from raw data.

'Deep' refers to depth — many stacked layers of computation between input and output. Each layer learns a slightly more abstract representation of the data than the one before it: early layers in an image model might detect edges, later layers detect shapes, and the last layers detect whole objects.