RNN Formulation & Execution Modes
Visualizing inputs, outputs, and temporal recurrence
🔤 Text/NLP
📈 Time Series
Autoregressive
Non-Autoregressive
Unfold Time Steps (View Outputs)
Run Data Flow
Folded Mode
U
V
W
X
H
O
Feeds back to next input
Feeds back to next input
U
V
W
X₁
hₜ₋₁
oₜ₋₁
U
V
W
X₂
hₜ
oₜ
U
V
X₃
hₜ₊₁
oₜ₊₁
Variable Breakdown
INPUT (xₜ)
--
HIDDEN STATE (hₜ)
Memory of past context
OUTPUT (oₜ)
--
Formulas
hₜ
= f(
U
·xₜ +
W
·hₜ₋₁ + b)
Memory
oₜ
= g(
V
·hₜ + c)
Prediction