luker501 / dposvoterrepresentation Goto Github PK

The Extended Justified Representation Axiom requires each committee to be assigned a score. Assigning a score to a committee naively requires calculating each individual voter score and then summing them up (this is what [this file does](https://github.com/Luker501/DPoSVoterRepresentation/blob/master/PAVScoreOfACommittee.java for the committee with the block producer candidates with indexes defined by the S array on line 14).

We have created a more efficient version of this algorithm, which we describe on the last page of this paper and we implement in this file.

But our efficient version of the algorithm does not return the same score for all of the committees it analyses. For example, if you use the example data provided, our efficient algorithm generates 3 separate committees, the last 2 of which have different scores compared to the naive implementation (see this file for further implementation).

This is mostly likely a rounding issue as we have many doubles and divisions in both files. But I have yet been able to resolve where the issue is, so that both files return the same score for every committee.