Penalize far apart pairs not more than as if they were two single ends #332
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #321.
The accuracy increases indeed a little bit.
Comparing accuracy
Average difference: +0.0004
I tried a couple of other things. In the formula you suggested in #321, you had an
epsilon_pair
term. I tested withepsilon_pair=2
andepsilon_pair=5
and while individual accuracies change a little bit for each dataset, the average difference remains at +0.0004 in both cases. I guess the term would have an effect for other datasets, so if you want, we can include it, but we canot measure its effect with the data we have.Also, as you suggested, I tried changing the three occurrences of the hard-coded -20 in the function to either -10 or -30 (this was in addition to introducing the
std::max()
). This made the average difference worse (-0.1860 and -0.0056, respectively).