Neuronika is a machine learning framework written in pure Rust, built with a focus on ease of use, fast prototyping and performance.

Dynamic neural networks and auto-differentiation

At the core of Neuronika lies a mechanism called reverse-mode automatic differentiation, that allows you to define dynamically changing neural networks with very low effort and no overhead by using a lean, fully imperative and define by run API.

The power of Rust

The Rust language allows for an intuitive, light and easy to use interface while achieving incredible performance. There's no need for a FFI, everything happens in front of your eyes.

Crate Feature Flags

The following crate feature flags are available. They configure the ndarray backend.

• serialize

  • Enables serialization support for serde 1.x.

• blas

  • Enables transparent BLAS support for matrix multiplication. Uses blas-src for pluggable backend, which needs to be configured separately. See here for more informations.

• matrixmultiply-threading

  • Enables the threading feature in the matrixmultiply package.

Project Status

Neuronika is very young and rapidly evolving, we are continously developing the project and breaking changes are expected during transitions from version to version. We adopt the newest stable Rust's features if we need them.