0
$\begingroup$

I am reading the paper of "Reasoning With Neural Tensor Networks for Knowledge Base Completion". I read it many times but I couldn't understand the parameters that are used especially the parameter U. they say that is one of the standard parameters of a basic neural net. I didn't get that, did I miss something or what?

The parameter U is not clear at all I search everywhere it stand for what but I could found!

$\endgroup$
1
$\begingroup$

I think you shouldn't read into it too much. It's simply a dot product between relation specific weights and the entity pair embedding. You can also think of $U$ as being the sole parameters in the last layer of the network which has no activation, no bias, and one output.

$\endgroup$
  • $\begingroup$ So you are telling me that I can eliminate it when I will use this formula again? $\endgroup$ – bttX Apr 30 '18 at 7:21
  • $\begingroup$ @bttX what do you mean "eliminate it" $\endgroup$ – shimao Apr 30 '18 at 7:27
  • $\begingroup$ I mean just rewrite the formula again without it :) $\endgroup$ – bttX Apr 30 '18 at 7:29
  • $\begingroup$ @bttX no, you can't just ignore it, because $g$ is supposed to be a scalar, and the output of $f$ is a vector... $\endgroup$ – shimao Apr 30 '18 at 7:32
  • $\begingroup$ maybe i can. i saw same formula but without this U like this one $\endgroup$ – bttX Apr 30 '18 at 7:35

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.