MATSim Code

use heuristic to generate starting points

add a collection of the distribution of ties weights during the primary process

delete commented out lines

allow to specify sampling rate for clustering estimation in config

use floats again...

still huge memory consumption

bugfix

store scores as doubles instead of float again

still store pointers as short, as exceeding the corresponding capacity
does not make sense. Float is much more problematic in terms of
precision and possibility of aver/underflow...

store scores as doubles instead of float again

hard limit on array size solves memory problems with limited impact on
results much better.

test maximal size behavior

re-introduce automatic scaling of storing array