nd4s

David did a great job with ScalaNLP. It happens to depend on Breeze. I would point out that Breeze does not support n-dimensional arrays (most tensors), although that is necessary to do in deep learning. We wrote ND4S and ScalNet to solve that: https://github.com/deeplearning4j/nd4s https://github.com/deeplearning4j/scalnet Moving computation out of Spark's MLlib and into lower level code like C++, as we do with JavaCPP and libnd4j, also improves speed. https://github.com/deeplearning4j/libnd4j https://github.com/bytedeco/javacpp

Yeah and it might be. We're targeting the big data folks with this framework frankly. Anything that does static typing might be overkill for just running matrix math. I created nd4j with the intent of having the same primitives and programming model accessible to me for production environments and also affording me the same privileges of optimizing things in c with nice clean abstractions. scala MIGHT be easier which is why we created nd4s: https://github.com/deeplearning4j/nd4s The JVM and the developers who use it are mainly people with salaries soliving a very specific class of problems not people publishing papers. The one exception to this I've seen is the NLP community. I'd argue it's just a different audience who needs the things the jvm has, eg: static typing, integrations with libraries, etc. For large codebases you really can't

The numpy array is a great data structure. I know it's not python but I wanted to plug our attempt at this for the jvm, nd4j: http://nd4j.org/tensor We have a java wrapper: https://github.com/deeplearning4j/nd4j Scala: https://github.com/deeplearning4j/nd4s And underlying c++: https://github.com/deeplearning4j/libnd4j We heavily dog food it in our deep learning library but we're hoping this can be more broadly useful for people.