how to use mean vector as default if the input node not in embedding layer

[liuxialong]

If the predict input node not in embedding layer.(for example the input node is out of bound)

I wanna use mean vector of the embedding layer as default.

How should I do?

[bkj]

In what scenario would you be feeding an "out of bound node" to the network? Do you mean a word at test time that was not in the training dataset?

[liuxialong]

@bkj

Yes. Maybe a word that not in the training dataset.

In that case, I may use mean vector as default in embedding layer.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs, but feel free to re-open it if needed.