<s>
A	O
recursive	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
is	O
a	O
kind	O
of	O
deep	O
neural	B-Architecture
network	I-Architecture
created	O
by	O
applying	O
the	O
same	O
set	O
of	O
weights	O
recursively	O
over	O
a	O
structured	O
input	O
,	O
to	O
produce	O
a	O
structured	B-General_Concept
prediction	I-General_Concept
over	O
variable-size	O
input	O
structures	O
,	O
or	O
a	O
scalar	O
prediction	O
on	O
it	O
,	O
by	O
traversing	O
a	O
given	O
structure	O
in	O
topological	B-Algorithm
order	I-Algorithm
.	O
</s>
<s>
Recursive	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
,	O
sometimes	O
abbreviated	O
as	O
RvNNs	O
,	O
have	O
been	O
successful	O
,	O
for	O
instance	O
,	O
in	O
learning	O
sequence	O
and	O
tree	O
structures	O
in	O
natural	B-Language
language	I-Language
processing	I-Language
,	O
mainly	O
phrase	O
and	O
sentence	O
continuous	O
representations	O
based	O
on	O
word	B-General_Concept
embedding	I-General_Concept
.	O
</s>
<s>
RecCC	O
is	O
a	O
constructive	O
neural	B-Architecture
network	I-Architecture
approach	O
to	O
deal	O
with	O
tree	O
domains	O
with	O
pioneering	O
applications	O
to	O
chemistry	O
and	O
extension	O
to	O
directed	O
acyclic	O
graphs	O
.	O
</s>
<s>
Recursive	O
neural	O
tensor	B-Device
networks	O
use	O
one	O
,	O
tensor-based	O
composition	O
function	O
for	O
all	O
nodes	O
in	O
the	O
tree	O
.	O
</s>
<s>
Typically	O
,	O
stochastic	B-Algorithm
gradient	I-Algorithm
descent	I-Algorithm
(	O
SGD	O
)	O
is	O
used	O
to	O
train	O
the	O
network	O
.	O
</s>
<s>
The	O
gradient	O
is	O
computed	O
using	O
backpropagation	B-Algorithm
through	I-Algorithm
structure	I-Algorithm
(	O
BPTS	O
)	O
,	O
a	O
variant	O
of	O
backpropagation	B-Algorithm
through	I-Algorithm
time	I-Algorithm
used	O
for	O
recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
.	O
</s>
<s>
Recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
are	O
recursive	O
artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
with	O
a	O
certain	O
structure	O
:	O
that	O
of	O
a	O
linear	O
chain	O
.	O
</s>
<s>
Whereas	O
recursive	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
operate	O
on	O
any	O
hierarchical	O
structure	O
,	O
combining	O
child	O
representations	O
into	O
parent	O
representations	O
,	O
recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
operate	O
on	O
the	O
linear	O
progression	O
of	O
time	O
,	O
combining	O
the	O
previous	O
time	O
step	O
and	O
a	O
hidden	O
representation	O
into	O
the	O
representation	O
for	O
the	O
current	O
time	O
step	O
.	O
</s>
<s>
An	O
efficient	O
approach	O
to	O
implement	O
recursive	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
is	O
given	O
by	O
the	O
Tree	O
Echo	O
State	O
Network	O
within	O
the	O
reservoir	B-Algorithm
computing	I-Algorithm
paradigm	O
.	O
</s>
<s>
Extensions	O
to	O
graphs	O
include	O
graph	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
(	O
GNN	O
)	O
,	O
Neural	B-Architecture
Network	I-Architecture
for	O
Graphs	O
(	O
NN4G	O
)	O
,	O
and	O
more	O
recently	O
convolutional	B-Architecture
neural	I-Architecture
networks	I-Architecture
for	O
graphs	O
.	O
</s>
