4 ms·

I wanted this article to convince me, but they really don't acknowledge the reason that everyone currently uses Python - the libraries. Do linear algebra using

I wanted this article to convince me, but they really don't acknowledge the reason that everyone currently uses Python - the libraries. Do linear algebra using lightly wrapped C? numpy. What about NLP? spaCy. Implicitly specify a computation graph with high level code? PyTorch. Explicitly specify a computation graph using leaky C++ abstractions? TensorFlow. And using either PyTorch or TensorFlow, you get to interact with CUDA. For now, if you want a functional language for doing deep learning, IMO that language needs Python interoperability. Long-term, I'm hopeful that GraalVM [0] can provide a way of calling Python from the JVM, but until then, I think the best option is coconut-lang [1], "a functional programming language that compiles to Python." You get pattern-matching, TCO, the pipe operator, and ADTs, all while being one AOT compilation step away from Python. [0] https://www.graalvm.org/ https://www.graalvm.org/ [1] http://coconut-lang.org/ http://coconut-lang.org/

bgorman 7y ago
For what it is worth, Apache MXNet has official Clojure bindings and you can use Deeplearning4j as well. There is also Neanderthal for CUDA/OpenCL programming in Clojure. Theoretically the ease of implementing DSLs in Clojure should be an advantage for the computation graph use case you are talking about.
zgramana 7y ago
If you haven’t checked it out already, you should take a look at Jython. As the name suggests, it’s a Java implementation of Python. https://www.jython.org/ https://www.jython.org/
- turingbike 7y ago
  I think the reason the parent mentions GraalVM instead of Jython is that Jython can't use C libs like numpy. But GraalVM lets you compile JVM to a native binary, so maybe there is some way to use Cython in JVM.
  - zgramana 7y ago
    JyNI* was created to bridge Jython to Numpy. I don’t have any personal experience with it, but one need not wait for a GraalVM solution. * https://www.jyni.org/ https://www.jyni.org/

bgorman 7y ago

For what it is worth, Apache MXNet has official Clojure bindings and you can use Deeplearning4j as well. There is also Neanderthal for CUDA/OpenCL programming in Clojure. Theoretically the ease of implementing DSLs in Clojure should be an advantage for the computation graph use case you are talking about.

zgramana 7y ago

If you haven’t checked it out already, you should take a look at Jython. As the name suggests, it’s a Java implementation of Python. https://www.jython.org/ https://www.jython.org/

turingbike 7y ago
I think the reason the parent mentions GraalVM instead of Jython is that Jython can't use C libs like numpy. But GraalVM lets you compile JVM to a native binary, so maybe there is some way to use Cython in JVM.
- zgramana 7y ago
  JyNI* was created to bridge Jython to Numpy. I don’t have any personal experience with it, but one need not wait for a GraalVM solution. * https://www.jyni.org/ https://www.jyni.org/