Fig. 4

This figure visualises the parallel summation of the thread projections in order to obtain the barycentre. In this example, three threads are randomly assigned to calculate the entire sum of all corresponding values \(p_{p,n}^{k,t}\), \(t=1,\ldots ,T\) to obtain \(p_{b,n}^k\). A racing condition is avoided since each thread operates on its own location in memory when saving \(p_{b,n}^k\)