2
$\begingroup$

I have a HP15-R203TX notebook with CPU i5-5200U 2.20GHzx4 / RAM 4Gb /2 Gb NVIDIA GeForce 820M. My friend has the same notebook. Another one has a dell notebook but same graphic card.

When i ran same jupyter notebook on same dataset to train the same number of layers, it takes 20 minutes on mine but 6-8 minutes on both of my friend's.

I tried resetting the whole environment, but didn't helped.

$\endgroup$
4
  • $\begingroup$ Wrong forum to ask this question. Perhaps try SO. $\endgroup$
    – DataD'oh
    Aug 11, 2017 at 9:33
  • $\begingroup$ SO would also get this question closed; I feel Super User is the best. $\endgroup$
    – Blaszard
    Aug 11, 2017 at 10:53
  • 1
    $\begingroup$ This question will be too broad almost anywhere. $\endgroup$
    – Stephen Rauch
    Aug 11, 2017 at 12:48
  • $\begingroup$ Even though your friend's PC have the same parameters, each can have different programs (mess) installed. This can have a big influence on the performance. $\endgroup$
    – HonzaB
    Aug 12, 2017 at 15:22

1 Answer 1

1
$\begingroup$

In machine learning, while your choice of inputs and hyper-parameters do matter, most of the time convergence ultimately comes down to how lucky you are. If you're lucky you might reach optimum in a very short period of time, whereas if you're unlucky you might get stuck in a local optimum with a huge error and get stuck in a very long loop.

That's one reason; another is possibly that if you're using TensorFlow and you installed the version for non-GPU computers, and your friends installed the version for computers with GPUs.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.