alpha
Login
or
Join now
gazagnaire.org
/
ocaml-ckms
Star
0
Fork
0
Atom
Configure Feed
Issues
Pull Requests
Commits
Tags
Feed URL
Select the types of activity you want to include in your feed.
Streaming quantiles with targeted error bounds (CKMS)
Star
0
Fork
0
Atom
Configure Feed
Issues
Pull Requests
Commits
Tags
Feed URL
Select the types of activity you want to include in your feed.
Overview
Issues
Pulls
Pipelines
ocaml-ckms
/
test
/
at
main
4 files
Thomas Gazagnaire
ckms: add CKMS streaming-quantile sketch
2mo ago
caa05dd6
dune
ckms: add CKMS streaming-quantile sketch A pure implementation of Cormode-Korn-Muthukrishnan-Srivastava biased quantiles (ICDE 2005): estimate chosen quantiles in memory bounded by the targets and their precision, not by the number of observations, with a guaranteed per-target rank error. Ckms.Sliding wraps it as a time-windowed estimator -- a ring of sketches rotated by an injected clock -- so a query reflects only recent observations, the shape a Prometheus summary needs. The clock is injected (unit -> float) to keep the core pure; tests drive it with a mock. Tested against the error-bound guarantee and the sliding-window decay.
2 months ago
test.ml
ckms: add CKMS streaming-quantile sketch A pure implementation of Cormode-Korn-Muthukrishnan-Srivastava biased quantiles (ICDE 2005): estimate chosen quantiles in memory bounded by the targets and their precision, not by the number of observations, with a guaranteed per-target rank error. Ckms.Sliding wraps it as a time-windowed estimator -- a ring of sketches rotated by an injected clock -- so a query reflects only recent observations, the shape a Prometheus summary needs. The clock is injected (unit -> float) to keep the core pure; tests drive it with a mock. Tested against the error-bound guarantee and the sliding-window decay.
2 months ago
test_ckms.ml
ckms: add CKMS streaming-quantile sketch A pure implementation of Cormode-Korn-Muthukrishnan-Srivastava biased quantiles (ICDE 2005): estimate chosen quantiles in memory bounded by the targets and their precision, not by the number of observations, with a guaranteed per-target rank error. Ckms.Sliding wraps it as a time-windowed estimator -- a ring of sketches rotated by an injected clock -- so a query reflects only recent observations, the shape a Prometheus summary needs. The clock is injected (unit -> float) to keep the core pure; tests drive it with a mock. Tested against the error-bound guarantee and the sliding-window decay.
2 months ago
test_ckms.mli
ckms: add CKMS streaming-quantile sketch A pure implementation of Cormode-Korn-Muthukrishnan-Srivastava biased quantiles (ICDE 2005): estimate chosen quantiles in memory bounded by the targets and their precision, not by the number of observations, with a guaranteed per-target rank error. Ckms.Sliding wraps it as a time-windowed estimator -- a ring of sketches rotated by an injected clock -- so a query reflects only recent observations, the shape a Prometheus summary needs. The clock is injected (unit -> float) to keep the core pure; tests drive it with a mock. Tested against the error-bound guarantee and the sliding-window decay.
2 months ago