Mlpack – a scalable C++ machine learning library

mlpack.org

88 points·michaelsbradley··8 comments

Open articleView on HN

mlpack was accepted into Google Summer of Code this year and there are some really cool projects going on right now:

https://summerofcode.withgoogle.com/organizations/5376684740...

See also the project's github: https://github.com/mlpack/mlpack

This is interesting, I like the way that each algorithm can be compiled to a command line executables.

However, the benchmark looks a bit confusing, how does it compare against some more popular libraries like tensorflow?

Also, while on black document is not very good for the eyes :)

How does Mlpack relate to tensorflow?

Tensorflow is aimed at ML architectures that are fully differentiable (just close your eyes and say "deep learning.") and performs automatic differentiation (you just define the architecture, tensorflow figures out how to update parameters). MLpack is less flexible but provides robust implementations of several class ML algorithms that you wouldn't necessarily want to write in Tensorflow, even though you probably could.

Does it have a distributed version?

Sorry if this sounds like a dumb question, can it be used for deep learning?

Is there an example of multiclass text classification with mlpack? I couldn't find any.

Not a user, but I saw this in the documentation: MulticlassClassificationLayer

Reply on news.ycombinator.com