Best OS and software for Machine Learning and Deep Learning

Which is the best OS for machine learning and deep learning

  • Mac OS

    Votes: 0 0.0%
  • Windows 7

    Votes: 0 0.0%
  • Others

    Votes: 0 0.0%

  • Total voters
    12

frozen

Member
Jul 29, 2017
36
0
6
30
Hello, I am a Mac user but I have also used Windows and Linux. I have a degree in AI many years ago but haven't worked on Machine Learning and Deep Learning. I am considering to replace my old computer and I am interested in learning Machine Learning and Deep Learning. Which is the best OS for Machine Learning and Deep Learning? What software do I need to install? Thank you.
 

Patrick

Administrator
Staff member
Dec 21, 2010
11,805
4,760
113
Ubuntu Linux w/ a NVIDIA GTX 1080 Ti is what you want as of today.

Then, get into nvidia-docker which will make life extraordinarily easy.
 
  • Like
Reactions: java and frozen

frozen

Member
Jul 29, 2017
36
0
6
30
Thanks Patrick. I only have an old MacBook Pro at home. Considering to get a MacBook Pro 2017 15" or a lightweight laptop with a desktop PC. Will that be good enough if I choose the latest MacBook Pro and add an eGPU GTX 1080 Ti? Should I choose an enclosure that would allow me to add multiple GTX 1080 Ti for future proof? I suppose it is better to install the Ubuntu Linux natively rather than though an emulator. Am I correct?

Not sure if Thunderbolt 3 and eGPU will work on the MBP under Ubuntu Linux.
 

Patrick

Administrator
Staff member
Dec 21, 2010
11,805
4,760
113
Thanks Patrick. I only have an old MacBook Pro at home. Considering to get a MacBook Pro 2017 15" or a lightweight laptop with a desktop PC. Will that be good enough if I choose the latest MacBook Pro and add an eGPU GTX 1080 Ti? Should I choose an enclosure that would allow me to add multiple GTX 1080 Ti for future proof? I suppose it is better to install the Ubuntu Linux natively rather than though an emulator. Am I correct?

Not sure if Thunderbolt 3 and eGPU will work on the MBP under Ubuntu Linux.
My advice instead of doing eGPU would be to buy/make a small server with a 1080 Ti. You can let that train models and have a reliable network connection while you access it remotely from the MBP. The eGPU enclosures are not that much cheaper than a used Xeon E3, E5-2670 (or E5-2660 V2), or low-end Xeon D server.

Think of it this way, if you end up wanting to train something that takes 2 hours, 8 hours or a day to run, do you want to leave your Mac attached the entire time? My sense is no.

The other advantage is that you will start learning to do ML on servers which is what you will do when you scale out to clusters or cloud instances.
 
  • Like
Reactions: frozen

frozen

Member
Jul 29, 2017
36
0
6
30
Thanks. I would assume that 4K display would allow us to see more especially when using TensorFlow. Is it advisable to get 4K display? I know that Windows has scaling problems. Not sure about Ubuntu. Any idea?
 

dhenzjhen

Member
Sep 14, 2016
38
55
18
San Jose, California
Thanks. I would assume that 4K display would allow us to see more especially when using TensorFlow. Is it advisable to get 4K display? I know that Windows has scaling problems. Not sure about Ubuntu. Any idea?
No need for fancy 4k monitor since you can do it remotely. Here's the run did last time just in case you want to see it using caffe and DIGITS.
 

Kritesh Anand

New Member
Dec 3, 2019
1
0
0
If you are using standard Machine Learning software packages like JMP, Weka, RapidMiner etc to perform basic operations like analysis, model creation etc, then Windows Operating system is a good choice.

However, Linux based Operating Systems are far more widely used for developing ML applications. Python, the undisputed king among the languages used for ML runs best in Linux where all dependencies can be installed with ease. Similar is the case for R and Octave, the other popular languages. Tensorflow, which has become one of the most powerful toolkits for Deep Learning runs best on Linux. There are a variety of cloud OS images that come pre-installed with the standard ML libraries -they are predominantly Linux based. You’ll end up with a generating graphs and other visualizations, there are a ton of them available for Linux.