desired outputs, predict outputs on future inputs.
fixed set, and scalar rewards/punishments, learn to select action sequences in a way that maximizes expected reward, e.g. chess and robotics. (This is more akin to learning how to design good experiments and is not covered in this course.)