In One Sentence

You trained a binary non-sparse RBF-C-SVM using libSVM, now you want the highest possible performance during (real-time) classification, like games or VR.

Highlights

• can load trained libSVM models (currently binary RBF-CSVM without sparse attributes)
• optimized for SIMD and can be mixed seamlessly with Rayon.
• allocation-free during classification
• written in 100% Rust, but can be loaded from any language (via FFI)
• 2.5x - 14x faster than libSVM

Principal Usage

Train with libSVM (e.g., using the tool svm-train), then classify with ffsvm-rust.

From Rust:

```rust // Get your libSVM model string from wherever and parse it. let modelstr: &str = includestr!("model.libsvm"); let model = ModelFile::tryfrom(modelstr).unwrap();

// Produce actual SVM from raw model, and a problem let csvm = RbfCSVM::try_from(&model).unwrap(); let mut problem = Problem::from(&csvm);

// Set the features of this problem we want to classify. problem.features = vec![ 0.3093766, 0.0, 0.0, 0.0, 0.0, 0.1764706, 0.1137485 ];

// (We also have methods to classify multiple in parallel w. Rayon ...) csvm.predict_value(&mut problem);

// Results should match libSVM assert_eq!(0, problem.label); ```

From C / FFI:

Please see FFSVM-FFI

Status

• Aug 5, 2018: Still in alpha, but finally on crates.io.
• May 27, 2018: We're in alpha. Successfully used internally on Windows, Mac, Android and Linux on various machines and devices. Once SIMD stabilizes and we can cross-compile to WASM we'll move to beta.
• December 16, 2017: We're in pre-alpha. It will probably not even work on your machine.

Performance

Classification time vs. libSVM.

Performance milestones during development.

See here for details.

FAQ

See here for details.