Module corpus 

Corpus Management for Online Learning

Provides efficient corpus storage with deduplication, importance sampling, and configurable eviction policies.

References

• [Vitter 1985] “Random Sampling with a Reservoir”
• [Settles 2009] “Active Learning Literature Survey”

Toyota Way Principles

• Muda Elimination: Deduplication avoids redundant training data
• Heijunka: Eviction policies level data quality over time

Structs

CorpusBuffer

Efficient corpus storage with deduplication

CorpusBufferConfig

Configuration for corpus buffer

CorpusMerger

Merge multiple data sources with configurable weighting

CorpusProvenance

Provenance tracking for merged corpus

CorpusSource

Source for corpus merger

Sample

A single sample in the corpus

Enums

EvictionPolicy

Eviction policy for corpus buffer

SampleSource

Sample source for provenance tracking