Why don't I see a GRU anywhere with more than one layer of perceptrons inside, it's pretty obvious to try to put more layers in there, but I don't see anyone doing that
3
-
$\begingroup$ Could you be more precise? There isn’t any “layer of perceptron” in GRU. $\endgroup$ – Tim♦ Jun 24 '20 at 18:06
-
$\begingroup$ I think he means "why are gates" computed using only a single linear layer instead of more complex multi-layer perceptrons. It's a good question... There are many works that use "deep-gated" recurrent units... For instance bmcbioinformatics.biomedcentral.com/articles/10.1186/… . $\endgroup$ – bomzh Jun 24 '20 at 18:40
-
1$\begingroup$ @bomzh it'd be nice if this was clarified, since there may be different interpretations of this question. $\endgroup$ – Tim♦ Jun 25 '20 at 13:07
Add a comment
|
