Target Pattern
0
0
0
0
Train
Test Input Sequence
0
0
0
0
0
Run Inference
→
↓
LSTM Cell (8 units)
x
t
(current bit) + h
t−1
(prev hidden)
↓
linear projections
Gates
⟠ computed in parallel
Forget f
t
σ(W
f
·[h,x]+b)
Input i
t
σ(W
i
·[h,x]+b)
Candidate g
t
tanh(W
g
·[h,x]+b)
Output o
t
σ(W
o
·[h,x]+b)
↓
f
t
⊙ c
t−1
+ i
t
⊙ g
t
Cell State c
t
↓
o
t
⊙ tanh(c
t
)
Hidden State h
t
↓
Dense(1) → σ
–
→
↓
Output
0