Module learned_index 

Learned vector indexes using neural networks

This module implements learned index structures that use neural networks to learn data distributions and provide faster lookups than traditional index structures.

Key Concepts

• Recursive Model Index (RMI): Hierarchy of models for indexing
• Learned CDF: Neural networks learn cumulative distribution function
• Error Bounds: Track prediction errors for correctness guarantees
• Hybrid Approach: Combine learned models with traditional search

References

• “The Case for Learned Index Structures” (Kraska et al., 2018)
• “Learning to Hash for Indexing Big Data” (Wang et al., 2016)

Re-exports

pub use config::LearnedIndexConfig;

pub use config::ModelArchitecture;

pub use config::TrainingConfig;

pub use neural_index::NeuralVectorIndex;

pub use rmi::RecursiveModelIndex;

pub use rmi::RmiStage;

pub use training::IndexTrainer;

pub use training::TrainingStats;

pub use types::LearnedIndexError;

pub use types::LearnedIndexResult;

pub use types::PredictionBounds;

Modules

config

Configuration for learned indexes

neural_index

Neural network-based learned index

rmi

Recursive Model Index (RMI) implementation

training

Training logic for learned indexes

types

Core types for learned indexes