Machine learning thread.
We had a lot of fun last time.
Let's discuss neural networks and neural network training accessories.
Should I learn Python or Scala for machine learning?
>>57918794
Assembly and python.
>>57918794
My personal suggestion would be python, because of keras, tensorflow and theano.
I literally started working with python because of NN libraries on it, I'm a perl users and have a great deal of negative bias against python - and so far I'm fairly satisfied with what I'm getting.
>>57918834
>Assembly
why
>>57918860
Because you can use python to call assembly.
>>57918900
And how good is that for working with neural networks..?
>>57918786
Actually kek'd at your pic OP, thanks.
>>57918910
https://prologsource.wordpress.com/2009/05/08/ultra-fast-hybrid-genetic-algorithm-in-assembly-language-for-the-travelling-salesman-problem-dll-for-lpa-prolog/
>>57918961
I don't get it. Are you suggesting to manually write optimizer for CPU in assembly, so that you can train your network at 1% of the speed that a GPU can provide?
>>57918986
Why can't you use gpu assembly?
>>57919014
Ask GPU vendors.
>>57919014
Different architectures
>>57919128
Yes, so?
>>57919492
Writing code in assembly to optimize your a neural network on GPU would be about as efficient use of your time as writing a distributed RDBMS in javascript, using only IE5 as the interpreter. It would be a tremendous waste of it.