In order to be able to evaluate results mult.info is required. This is
produced by JumboDBG in a run when we are creating ML graph samples.
With new interface, IDs are not stable anymore so we need to use GFA
files to hash the edge strings and create mapping between JumboDBG outputs.
This PR takes care of multiple things:
Restore gfa parsing with new interface
Adds idmap to kept files
Adds hash mapping of the ids for evaluation
Ensure hashmap is created during the test sampling
Update JumboDBG output to have RC included in edge id
Map ids before confusion matrix calculation
It seems that LJA runs undeterministically so it might happen that couple of
edges are correct in one run and incorrect in another run, but this does > [!NOTE]
affect overall results due to the large number of edges.
In order to be able to evaluate results mult.info is required. This is produced by JumboDBG in a run when we are creating ML graph samples. With new interface, IDs are not stable anymore so we need to use GFA files to hash the edge strings and create mapping between JumboDBG outputs.
This PR takes care of multiple things:
It seems that LJA runs undeterministically so it might happen that couple of edges are correct in one run and incorrect in another run, but this does > [!NOTE] affect overall results due to the large number of edges.