Closed prernakhurana2 closed 7 years ago
Actually I can't replicate what the author did. Attention weights it learned are pretty much equally weighted. I am wondering if any trick such as adversarial perturbation on the word vector layer can help.
Actually I can't replicate what the author did. Attention weights it learned are pretty much equally weighted. I am wondering if any trick such as adversarial perturbation on the word vector layer can help.