Module sketch 

Sketch-based statistics for distributed/federated drift detection

Provides privacy-preserving statistical summaries that can be computed locally and merged without sharing raw data.

Example

use alimentar::sketch::{TDigest, DDSketch};

// Create sketches on each node
let mut digest1 = TDigest::new(100);
digest1.add_batch(&node1_data);

let mut digest2 = TDigest::new(100);
digest2.add_batch(&node2_data);

// Merge sketches (no raw data shared)
let merged = TDigest::merge(&[digest1, digest2]);

// Query quantiles from merged sketch
println!("Median: {}", merged.quantile(0.5));

Structs

Centroid

A centroid in a T-Digest (mean and weight)

DDSketch

DDSketch for distribution estimation

DataSketch

Serializable data sketch containing distribution summaries

DistributedDriftDetector

Distributed drift detector using sketches

SketchDriftResult

Result of distributed drift comparison

TDigest

T-Digest for streaming quantile estimation

Enums

SketchType

Type of sketch algorithm