89 points by rust_ml_lover 1 year ago flag hide 10 comments
rust_ml_creator 1 year ago next
Excited to share my new Machine Learning library in Rust! It's been a long journey and would love to hear feedback from the HN community.
learned_rust 1 year ago next
Congrats on the release! I've been learning Rust recently and am curious about the challenges you faced building a ML lib in Rust as opposed to some other languages.
rust_ml_creator 1 year ago next
Thank you! One challenge was efficiently interfacing between Rust and the underlying linear algebra libraries. Since Rust doesn't have built-in linear algebra functionality, interfacing with libraries like BLAS/LAPACK required some additional effort.
haskell_lover 1 year ago prev next
Are there any benefits of using Rust for ML over Haskell?
functional_fan 1 year ago next
While both have their merits, Rust offers more consistent performance while Haskell might be a bit slower due to its garbage collection. However, Haskell's purity could result in more predictable code and fewer bugs.
numerical_wizard 1 year ago prev next
Impressive work! Which numerical optimizations did you implement and which algorithmic improvements are planned for future releases?
rust_ml_creator 1 year ago next
Thanks! We've implemented some basic SIMD optimizations for vector operations and used blocked algorithms for linear algebra. As for future improvements, we're looking to implement automatic differentiation and GPU support.
pythonmlfan 1 year ago prev next
What about Python integration? Any plans for that?
rust_ml_creator 1 year ago next
We have plans to support Python bindings through projects like PyO3 or rust-bindgen. Our goal is to make sure our library is accessible to a wide range of ML developers.
pyo3_maintainer 1 year ago next
Looking forward to your contributions! PyO3 is designed to facilitate and streamline the creation of Python extensions within Rust, making it a perfect choice for your needs.