How to compare 2 sets (vectors), features, without normalization?

Question

Algorithms

I want to compare 2 image invariants presented as sets M1 .... M7
There are 2 vectors of numbers, you need to compare them, and display the percentage of similarity. For example, a1 (10, 3, 9999999, 0.00000001), a2 (11, 3, 9999999, 4444). From the example, the 4th value stands out a billion times, but the other 3 are approximately equal.
I know there are some distances, but even I don’t know how to normalize.
A solution option is that each feature has an equal maximum percentage of similarity (from the example for 4 numbers of 25% each). 75% comes out of the example.
But anyway, I don't know how to evaluate, without normalization. For example, the number 10 and 11 is how much different, and 10 and 20, or 10 and 100000.