Repository navigation
Read the permutation null's draws from a rank table when every block ranks against its whole bucket - #338
Merged
Conversation
…ranks against its whole bucket A uint16 table of each key's rank within its profile's sorted row replaces the drawn-pair similarities and the searches, so a draw scores nothing. It is used when it fits RANK_BUDGET; otherwise the sorted keys serve as before. _tables also returns early when it can table nothing.
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #338 +/- ##
==========================================
- Coverage 76.48% 76.16% -0.32%
==========================================
Files 100 100
Lines 13813 13934 +121
==========================================
+ Hits 10565 10613 +48
- Misses 3248 3321 +73
🚀 New features to boost your workflow:
|
…t.settings.map_table_bytes One setting, half of the machine's memory by default, replaces the fixed 8 GiB store budget and the rank-table budget. A rank table that does not fit beside the sorted similarities is logged and skipped.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When every block ranks against its whole bucket with nothing subtracted (replicability and cross_plate modes), the permutation null now builds a uint16 table of each key's rank within its profile's sorted row. A draw then counts the negatives before a positive as its rank less the ranks of the drawn profiles below it, so it scores no similarities and searches no rows.
The table is used only when it fits beside the sorted similarities in the new
mt.settings.map_table_bytes(half of the machine's memory by default, replacing the fixed 8 GiB store budget), rows hold at most 65,535 keys and no key is NaN. Activity mode, multilabel columns and larger tables keep the current path, and a table skipped for its size is logged._tablesalso returns early when it can table nothing.Verified: a new test checks that the rank table and the sorted keys give identical p-values and scores, with and without ties and duplicate profiles, and that a budget below the table size falls back. The copairs equivalence and tl evaluation tests pass (185). On Aconcagua replicability at null_size 10,000, the whole scoring run takes about 8.5 minutes.
On Aconcagua (main's r7 file, replicability, null_size 10,000), scores, p-values and q-values are bit-identical to main for all 500 groups.