Increasing the efficiency of a codification for a sequence predictor has a fundamental role and the size of the resulting code is the most important characteristic because the prediction of a sequence require a time exponential on the size of the code .
Using a code without redundancy for combinatorial data can reduce a lot the size of the code so searching how to code combinatorial data I found a very interesting algorithm the Combinadic .
This algorithm allow to construct a correspondence for every combination ( N K ) in a natural number 0…N!/(K!*(N-K)! ) -1 and viceversa with a polynomial time complexity
Here a link for the code