Lucas Fidon edited subsubsection_Acceleration_based_metric_Contrary__.tex  almost 8 years ago

Commit id: 0282eba222c9fadfb1cce521cfac37ab07ec5fbb

deletions | additions      

       

Basically we get one big cluster embedding a lot of players and several 1-player cluster. The biggest cluster behaves like a "sink cluster": as we increase the value of $C_{penalty}$ the 1-player clusters tend to be directly absorbed by the biggest cluster.  As a consequence there is no telling about nothing can be infer from  the results of this acceleration-based distance. It does not allow to solve our problem.