docs(eval): per-family accuracy (all 6 families, real gct) - #492
Merged
Merged
Conversation
Adds Fig 3(c): Sum exact, DDSketch/KLL per-series quantiles (medians in-envelope, p99 tail = small-N collapse variance), CMS exact freq, HLL per-series 3e-5 — 5/6 families validated. DDSketch-vs-KLL head-to-head (KLL better median, DDSketch tighter tail + -26% wire). Two named gaps: CountSketch topk recall-0 (keys by item not host, ranks by occurrence not value → value-weighted topk needs a separate path) and HLL global rollup (per-series exact, global not served). Notes wall-clock anchoring was the warm-read fix (timing not reducer) + the warm-vs-archive range-routing finding.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Folds the multi-sketch per-family accuracy into Fig 3(c), on real gct (1000 series): Sum exact, DDSketch/KLL quantiles in-envelope, CMS exact freq, HLL per-series 3e-5 — 5/6 families validated. DDSketch-vs-KLL head-to-head (KLL better median; DDSketch tighter tail, −26% wire). Two named gaps: CountSketch topk recall 0 (keys by item-not-host, ranks by occurrence-not-value → value-weighted topk needs a separate path) and HLL global rollup. Notes wall-clock anchoring fixed the warm read (timing, not reducer) + the warm-vs-archive range-routing finding. Data:
datasets_eval/multisketch/(feat/multisketch-eval + feat/multisketch-accuracy).🤖 Generated with Claude Code