https://dragan.rocks/articles/18/Neanderthal-vs-ND4J-vol2

If you know your way with Python and Numpy, a direct head-to head comparison will be a nice advertisment for Clojure. I simply don't have time to fiddle with Python, but will be delighted to do the Neanderthal part and the writeup!

@blueberry I’m not a python person either but maybe a script like this could help? https://gist.github.com/markus-beuckelmann/8bc25531b11158431a5b09a45abd6276

@gigasquid Maybe we could also check Neanderthal vs MXNet's backend?

Yes, I could hack away something, but I would like to give Numpy a fair chance.

MXNet is all NDArray

I know.

It does the same ops though.

oh you mean benchmark

Yes

So, the Numpy part has to be written by someone who actually uses it.

I’ll keep that in mind - we don’t have any benchmarks yet for the Clojure MXNet package, but I have it as an item that needs contribution help

https://cwiki.apache.org/confluence/display/MXNET/Clojure+Package+Contribution+Needs

It would be better if the benchmark uses vanilla Java/Scala MXNet via interop, to avoid complaints that MXNet was crippled by Clojure's wrapper.

heh

I’ll look around and see if there is some benchmark code especially on the Scala side because I’m interested the difference with the clojure-scala interop

There is no pure Java API