Rusty-machine has just reached version 0.3!
What is rusty-machine?
Rusty-machine is a general purpose machine learning library written entirely in Rust. I wanted to create a library which could support both performance and ease of use – without a large number of external dependencies. Rust lets us write a modular library with a simple interface which can be easily extended by the user. For a more concrete explanation check out my previous blog post .
A lot has changed since v0.2. Probably too much to list everything so I’ll stick with the important stuff.
A lot of work went into optimizing the linear algebra. It’s no longer terrible (but still not state of the art). Thanks to matrixmultiply we now have a fairly competitive matrix multiplication implementation. Along the same lines we’ve also introduced the idea of a
MatrixSlice . This behaves like a slice in the standard library but for matrices! It has already helped us remove a lot of copying and allocation – with more to come.
There have also been some additions to the learning module. We’ve introduce two new machine learning algorithms: Naive Bayes and GMMs . Both follow the same pattern as the other models – focusing on modularity and customizability.
To learn more, read the full changelog .