There is a huge space leak that prevents the algorithm from working properly when the number of centers is big (> ~ 5000). The deepseq branch solves this, but at the expense of a significant performance cost.
Most likely, using more appropriate data structures will solve the issue. However, further understanding the leak will probably shed some light on which data structure would be best suited for this implementation.
There is a huge space leak that prevents the algorithm from working properly when the number of centers is big (> ~ 5000). The deepseq branch solves this, but at the expense of a significant performance cost.
Most likely, using more appropriate data structures will solve the issue. However, further understanding the leak will probably shed some light on which data structure would be best suited for this implementation.