aprender 0.27.7

A pure Rust machine learning library with SIMD acceleration, GPU inference, and the APR v2 model format. No Python dependencies, no C bindings -- memory-safe, thread-safe, and WebAssembly-ready.

Features | Installation | Quick Start | Algorithms | APR Format | Architecture | Quality | Sovereign Stack | Documentation | License

Table of Contents

• Features
• Installation
• Quick Start
• Algorithms
• APR Format
• Architecture
• Quality
• Sovereign AI Stack
• Documentation
• Contributing
• License

What is aprender?

Aprender is a production-ready machine learning library written entirely in Rust. It provides classical and modern ML algorithms -- from linear regression and k-means to neural networks and transformers -- with SIMD acceleration via trueno and GPU inference via realizar. Models serialize to the APR v2 format with LZ4/ZSTD compression, zero-copy loading, and optional AES-256-GCM encryption.

Aprender is the ML foundation of the PAIML Sovereign AI Stack.

Features

• Pure Rust -- Zero C/C++ dependencies, memory-safe and thread-safe by default.
• SIMD Acceleration -- Vectorized operations via trueno 0.16 (AVX2/AVX-512/NEON).
• GPU Inference -- CUDA-accelerated inference via realizar (67.8 tok/s 7B, 851 tok/s 1.5B batched on RTX 4090).
• APR v2 Format -- Native model serialization with LZ4/ZSTD compression, zero-copy loading, and Int4/Int8 quantization.
• Multi-Format -- Native .apr, SafeTensors (single + sharded), and GGUF support.
• WebAssembly Ready -- Compile to WASM for browser and edge deployment.
• 12,974+ Tests -- 96.35% coverage, 73 provable contracts.

Installation

Add aprender to your Cargo.toml:

[dependencies]
aprender = "0.27"

Optional Features

[dependencies]
aprender = { version = "0.27", features = ["format-encryption", "hf-hub-integration"] }

Quick Start

use aprender::prelude::*;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Training data
    let x = Matrix::from_vec(4, 2, vec![
        1.0, 2.0,
        2.0, 3.0,
        3.0, 4.0,
        4.0, 5.0,
    ])?;
    let y = Vector::from_slice(&[3.0, 5.0, 7.0, 9.0]);

    // Train model
    let mut model = LinearRegression::new();
    model.fit(&x, &y)?;

    // Evaluate
    println!("R-squared = {:.4}", model.score(&x, &y));

    Ok(())
}

Algorithms

Linear Models

Clustering

Classification

Decision Trees

NLP and Text

Time Series

Graph Analysis

Bayesian Inference

Neural Networks

Decomposition

Model Serialization (APR v2)

Online Learning

Recommendation Systems

APR Format

The .apr format provides secure, efficient model serialization:

use aprender::format::{save, load, ModelType, SaveOptions};

// Save with encryption
save(&model, ModelType::LinearRegression, "model.apr",
    SaveOptions::default()
        .with_encryption("password")
        .with_compression(true))?;

// Load
let model: LinearRegression = load("model.apr", ModelType::LinearRegression)?;

Architecture

aprender
  primitives/       Core tensor types (Matrix, Vector)
  linear_model/     Linear and logistic regression
  cluster/          KMeans, DBSCAN
  classification/   SVM, Naive Bayes, KNN
  tree/             Decision trees
  ensemble/         Random forest, gradient boosting
  text/             Tokenization, TF-IDF, chat templates
  time_series/      ARIMA
  graph/            PageRank, centrality, community detection
  bayesian/         Bayesian inference
  glm/              Generalized linear models
  decomposition/    PCA, dimensionality reduction
  nn/               Neural networks, transformers, MoE
  online/           Online / incremental learning
  recommend/        Recommendation systems
  synthetic/        Synthetic data generation
  format/           APR v2 serialization (encryption, compression)
  serialization/    Legacy serialization
  prelude/          Convenient re-exports

Key dependency: trueno provides the SIMD-accelerated compute primitives (AVX2/AVX-512/NEON).

Quality

• 12,974+ tests, 96.35% line coverage
• Zero clippy warnings (-D warnings)
• 73 provable contracts via provable-contracts
• TDG Score: A+ (95.2/100)
• Mutation testing target: >80% mutation score
• Zero SATD (self-admitted technical debt)

Sovereign AI Stack

Aprender is the ML layer of the PAIML Sovereign AI Stack -- a pure Rust ecosystem for privacy-preserving ML infrastructure.

Related Crates

Documentation

Contributing

1. Fork the repository
2. Make your changes on the master branch
3. Run quality gates: cargo test --all-features && cargo clippy --all-targets -- -D warnings && cargo fmt --check
4. Submit a pull request

See CONTRIBUTING.md for guidelines.

License

MIT