This pull request consists of updated results, after accounting for an index mismatch bug I found in the code. Candidates are extracted from the database in sorted order (by their ids); however, when I created an excel file to look at the development set's candidates, I shuffled the order. Since I didn't account for this, the ROCs and the PRs showed poor performance. Now that I have remedied this issue, some of the results are looking a bit better. Will add the discriminator performance in the next pull request.
This pull request consists of updated results, after accounting for an index mismatch bug I found in the code. Candidates are extracted from the database in sorted order (by their ids); however, when I created an excel file to look at the development set's candidates, I shuffled the order. Since I didn't account for this, the ROCs and the PRs showed poor performance. Now that I have remedied this issue, some of the results are looking a bit better. Will add the discriminator performance in the next pull request.