1
$\begingroup$

I'm trying to build a neural network between protein sequence and its drug fingerprint. My input size is 20000. The output size is 881. The sample size is 610.

Can I process this huge neural network? But how? And in which tool?

$\endgroup$
1
  • $\begingroup$ it seems sort data sample to adjust a NN and get good predictions from it...when you adjust and try please let us know if you have got useful trained patterns from this data. Also, take into account that NN recognize patterns, I doubt you can find a correlation between slices in your proteins and drug fingerprint slices. $\endgroup$ – Rogelio Triviño Aug 27 '20 at 15:33
2
$\begingroup$

Yes, it should be no problem.

When you decide to use a CNN, you have to make sure that this makes sense. Another answer mentioned using 3x3 convolutions -- which I would recommend against. For that to work, you would need to turn your vector into a rectangular array, and you would be implying a structure that isn't there.

Use one-dimensional convolutions instead.

$\endgroup$
0
$\begingroup$

It sure is possible, imagine a CNN can handle way bigger number of inputs. An image with size of 512x512 has already 262144 input nodes when re-arranged to a one-row vector. The trick sicne 2012/2014 is to use Convolutions, and deep ones, so stacking a lot of 3x3 Convolutions for example. Its way less sensitive than a fully-connected Dense network and needs a siginificant amount of less parameters. For more check this out, chapter 9 : Ian-Goodfellow, Deep Learning

Tools for that are tensorflow and keras based on python, or tensorflow-js on java, you can also use pytorch but the community is rather small on comparison.

$\endgroup$
1
  • $\begingroup$ The 3x3 convolution only makes sense for data that is inherently two-dimensional. When you fold a one-dimensional vector into a two-dimensional array, you are imposing a structure that is not really there: two points that are vertical neighbors don't necessarily have any relationship to each other; only horizontal neighborhoods matter. $\endgroup$ – Robby Goetschalckx Aug 27 '20 at 17:30

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.