5
(Choose 1 answer)
Here are the equations for the GRU and the LSTM:
GRU
LSTM
<t> = tanh(W[Tr, c<t-1>, x<t>] + bc)
<t> = tanh(We[a<t-1>, x<t>] +bc)
Tu = (Wulc<-1>, x<t>] + bu)
Γ = (Wula<t-1>, x<t>] + bu)
Γ, = (W,[c<-1>, x<t>] + br)
Γ = (W[a<-1>, x<t>] + b)
c<t> = u č<t> + (1) * c<t-1>
Γο= σ(Wo[a<-1>, x<t>] + bo)
a<t> = c<t>
A. C
B. A
C. D
D. B
c<t> = Tu * <t> + [ * c<t-1>
a<t> = To * c<t>
From these, we can see that the Update Gate and Forget Gate in the LSTM play a role similar to and in the GRU. What should go in the the blanks?
a.Γ and 1 - Γ
b. T and I,
c. 1- and
d. I, and
