0
$\begingroup$

I have a user friendship graph for about 30 Million users. I am trying to use an auto encoder [30 Million, 512, 512, 1024, Dropout(0.3), 512, 512, 30 Million]. But I am not learning anything. Has anyone used Autoencoder for Friendship recommendation on a large scale graph?

$\endgroup$
0
$\begingroup$

Autoencoder will not predict potential friendship. Autoencoders learn a lossy compression of the data. A lossy compression of existing friendship relationships will not predict new relationships.

|improve this answer|||||
$\endgroup$
  • $\begingroup$ How do you do link prediction with deep learning? How do you use collaborative filtering for friendship recommendations? $\endgroup$ – Veshnu Ramakrishnan May 18 '19 at 3:01
  • $\begingroup$ Those are independent questions. Ask them as independent questions, not comments. $\endgroup$ – Brian Spiering May 18 '19 at 14:26

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.