Sequence Processing Tasks
Recurrent Neural Networks
x1
x2
x3
x4
x5
U
U
U
U
U
W
W
W
W
V
V
V
V
V
s1
s2
s3
s4
s5
y^1
y^2
y^3
y^4
y^5
s0
W
Find a cheap Chinese restaurant
si=RNN(si−1,xi)
VB DT JJ JJ NN
Recurrent Neural Networks
xi
U
st
y^i
st−1
W
V∈R36×d
HMM vs RNN
HMMs are simpler than RNN
HMMs have less parameters, hence require less data
HMM vs RNN
P(yt∣yt−1,...,y2,y1)=P(yt∣yt−1)
P(w6∣w5,w4,..,w1)
x1
x2
x3
x4
x5
U
U
U
U
U
W
W
W
W
V
s1
s2
s3
s4
s5
y^5
s0
W
Find me a cheap Chinese
HMM vs RNN
VB
DT
JJ
Find
a
cheap
P(x)=∑yP(x,y)
=∑yP(x∣y)P(y)
=∑y∏tP(xt∣yt)∏tP(yt∣yt−1)
HMM vs RNN
x1
x2
x3
x4
x5
U
U
U
U
U
W
W
W
W
V
s1
s2
s3
s4
s5
y^5
s0
W
P(y^5∣x5,x4,..,x1)