Closed aldro61 closed 7 years ago
I wonder if we should let the user specify the numerical tolerance parameter at run time? (rather than specifying it at compile time) I guess we can always make that change later on, if we run into any problems on other data sets.
for now I think this is ok to merge with master.
The breakpoint list was still not using 6 decimal precision, leading to solver errors in some rare cases. This has been corrected. No more solver errors when learning from the neuroblastoma data.