0
$\begingroup$

I have a large number of x86_64 cores available to me, but no GPUs or TPUs. Which Machine Learning techniques are suited for execution on a CPU? I would imagine more "statistical learning" techniques rather than "deep learning".

The reason I ask is that I'm interesting in studying some state-of-the art techniques, but I'm in the situation noted above.

$\endgroup$
1
$\begingroup$

With 12 or more CPU cores, you wan work with pretty much any deep learning model. Only thing that will not work is training with 100k+ images or more than TBs of test. For such scenarios; you can seed the model with transfer learning.

For example, training a Cat/Dog image classification model (from scratch) takes 2 hours on 6 core i5 (as compared to 30 minutes on GTX 1080ti).

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.