What would be still affordable consumer tier hardware, but strong enough to train mediocre neural networks?
If anyone has experience with machine Learning, just how hardware intensive is it?
I started to get into it recently and trained some simpler networks, but now i want more.
>>58690072
How good a programmer are you? How's your math?
>>58690912
Both is decent, but i dont have too much real world experience yet.
Went through math and CS on a prestigious university.
But you dont really need strong math skills to implement machine learning with the correct frameworks right?
I mean all the algorithms are already there, you just have to design the network correctly and adapt it to your problem and even then thats more finetunning/trying different configurations than math.
>>58690072
Depends on what you are doing, how deep your network is etc. You can get a lot of shit done with a modern GPU with 4GB vram. But if you are doing anything serious, you'll need 16GB or more.
>>58691862
Which task would you classify as serious?
>>58691452
That's very ignorant.
> Went through math and CS on a prestigious university.
Sure you did.
>>58692935
Well i did.
But feel free to correct me if i said anything ignorant, i am not trying to offend.
If you have legitimate advice i will gladly take it since i am a newbie in the ML field.
I do think i understand all the math behind the gradient optimiziers and neural nets in general though.
>>58692027
Tasks with big data. Like classification of cancer cases with volumetric DICOM data. You'll be hard pressed to make any serious models with just 4GB then. It can still be done probably, but you are on the edge of what is feasible.
>>58692935
>university is for fags, selftaught masterrace r-right
>>58690072
Look for the PC-building threads here and ask around for help. First set your budget, then prioritize the rest.
You could easily get an i5 2.7Mhz, 16 Gb RAM, 500 SSD + 500 HDD (this last one entirely optional) for maybe $600.