Today python has so many excellent libraries for ML, the ones we are checking with,
1. sklearn: http://scikit-learn.org/stable/ 2. Keras 2: https://blog.keras.io/introducing-keras-2.html?t=1 3. Theano: http://deeplearning.net/software/theano/ 4. Pytorch: https://github.com/pytorch/pytorch I do wish perl community has got involved ML world too. :) 2017-02-12 2:54 GMT+08:00 James Alton <jamesalton...@gmail.com>: > If you want to learn about machine learning and the first thing you think > is: "What Perl libraries are there for this?", you might be going about it > all wrong. > > *The best machine learning libraries are all written for Python*. This > isn't a "religious" argument, it's just the way it is. I always think: > Right tool for the job. In this case, it's Python. > > TensorFlow - C++ front end, Python front end > > Torch - Lua, Python wrapper > Theano - Python > Caffe - C++, Python wrapper > > (I put TensorFlow by itself because it's probably the one you want anyway. > Probably the best library out there right now for the vast majority of ML.) > > Another thing that follows this same theme with machine learning is: AMD > or Nvidia? Normally I'd say use what you've got, but in this case, it would > be the wrong decision. You need to go Nvidia otherwise you'll be a "second > class citizen" in the machine learning world. (OpenCL support often seems > like an after thought, CUDA is first.) And it has to be a fairly recent > Nvidia, otherwise stuff won't even load. (A year or two old is ok, point is > - research the limitation here carefully.) > > Your best bet from a Perl standpoint is: Do the heavy lifting in Python, > pass your results to Perl using JSON or something, and be done with it. > > *And btw, if you want the fast track on ML, try Keras.* (It's also for > Python.) It's an even higher level than Tensorflow and it's almost "fun" to > program ML in. (Hides a lot of the ugly bits. And with ML, there really are > a lot of ugly bits. The interface in Keras is like: train this model with > these layers, give me results with that trained model.) > > YMMV and I hope my experiences above saves you lots of potential wasted > time and money, > James > > > > > On Sat, Feb 11, 2017 at 11:28 AM, chace <chacewe...@gmail.com> wrote: > >> Hi there. >> >> I don't know much about machine learning myself, much less doing it in >> Perl. But I have dabbled and have been meaning to do a bit more with it in >> the Perl-based scientific computing framework PDL. >> >> A quick Google search revealed a short blurb on PerlMonks about ML >> analysis of stock data using PDL. The very last comment also points to a >> couple of AI modules on CPAN. Links here: >> >> http://www.perlmonks.org/?node_id=638391 >> >> http://search.cpan.org/search?mode=module&query=AI%3A%3ACategorizer >> >> http://search.cpan.org/search?mode=module&query=AI%3A%3ADecisionTree >> >> >> Perl-land is a bit less frameworky and more library-ish than Python-land. >> There's plenty of algorithms on the CPAN, so I suggest figuring out what >> you need and importing as you figure that out. >> >> >> On 02/07/2017 05:47 PM, community tech wrote: >> >>> Is there a popular perl library/framework for machine learning? >>> >>> Thanks. >>> >> >> -- >> To unsubscribe, e-mail: beginners-unsubscr...@perl.org >> For additional commands, e-mail: beginners-h...@perl.org >> http://learn.perl.org/ >> >> >> >
