Description Usage Arguments Value
Applies zoneout as described in Krueger et al. 2017
1 | zoneout(thisState, prevState, p)
|
thisState |
the current state (mx.symbol) |
prevState |
the state from the previous sequence (mx.symbol). |
p |
zoneout probability in (0;1]. Rounded to 6 digits |
During training: out(t) = x(t) * thisState(t) + (1 - x(t)) * prevState(t). x(t) is a 0/1 mask for each element t with the probability p of being 0. During application: out(t) = (1 - p) * thisState(t) + p * prevState(t)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.