MIT Press

Table 1:

Composition functions from the literature. u and v are the word representations of the inputs. The results reported in Section 5 use the identity function g(x) = x instead of a nonlinearity because no improvements were observed for these models using

g = tanh

or g = ReLU.

Name	Composition function f
Addition (Mitchell and Lapata, 2010)	u + v
SAddition (Mitchell and Lapata, 2010)	αu + βv
VAddition	a ⊙u + b ⊙v
Matrix (Socher et al., 2010)	g(W[u; v] + b)
FullLex (Socher et al., 2012)	g(W[A^vu;A^uv] + b)
BiLinear (Socher et al., 2013a,b)	$g (u^{⊺} E v + W [u; v] + b)$
WMask (Dima, 2015)	g(W[u ⊙u^m; v ⊙v^h] + b)

Name	Composition function f
Addition (Mitchell and Lapata, 2010)	u + v
SAddition (Mitchell and Lapata, 2010)	αu + βv
VAddition	a ⊙u + b ⊙v
Matrix (Socher et al., 2010)	g(W[u; v] + b)
FullLex (Socher et al., 2012)	g(W[A^vu;A^uv] + b)
BiLinear (Socher et al., 2013a,b)	$g (u^{⊺} E v + W [u; v] + b)$
WMask (Dima, 2015)	g(W[u ⊙u^m; v ⊙v^h] + b)

View Large

This Feature Is Available To Subscribers Only