mixt - Infer mixture model parameters

mixt provides algorithms for estimating the mixing proportions for a mixture model with a fixed log-likelihood matrix, ie. a model where the hyperparameters of the emission distributions are known but the weights are not.

Documentation is available at https://docs.rs/mixt.

Usage

mixt supports three main use cases:

• Rust library API.
• C++ API.
• Minimal CLI.

Both the Rust and C++ API support GPU acceleration with the Wgpu backend from burn.

About

mixt is a Rust rewrite of a previous C++ implementation. You can find the C++ code at tmaklin/rcgpar.

The C++ and Rust code have equal CPU performance, but the Rust code implements multiple numerical stability tricks that enable running on 32-bit floating point numbers, allowing for better GPU utilization.

burn allows compiling mixt for many different GPU architectures, whereas the C++ implementation only supports torch.

License

The source code from this project is subject to the terms of the LGPL-2.1 license. A copy of the LGPL-2.1 license is supplied with the project, or can be obtained at https://opensource.org/licenses/LGPL-2.1.