The concept of distance is an essential one in almost every field of engineering. While distance for most means the same thing, in reality there are an infinite ways to measure distance. These measure are known as norms in mathematics. What we use in our everyday life is known as Euclidean or L2 norm. L2 norm is immensely popular (heard of least square regression?) in research. But as time passes by we have explored a lot of different norms which outperform L2 in many tasks by a huge margin but for historic reason L2 still remains most widely used. L2's popularity is mainly because most computations involving it have analytical solutions and even with great advances in numerical methods we still feel comfortable when we have analytical solutions. This trend has to change....Its time we say no to L2
Its time we say no to L2
BTW: L1 is slowly gaining popularity with almost half a dozen papers published on it every month.
Print This Post