Skip to Main Content
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.
NameComposition function f
Addition (Mitchell and Lapata, 2010) u + v 
SAddition (Mitchell and Lapata, 2010) αu + βv 
VAddition au + bv 
Matrix (Socher et al., 2010) g(W[u; v] + b
FullLex (Socher et al., 2012) g(W[Avu;Auv] + b
BiLinear (Socher et al., 2013a,b) g(uEv+W[u;v]+b) 
WMask (Dima, 2015) g(W[uum; vvh] + b
NameComposition function f
Addition (Mitchell and Lapata, 2010) u + v 
SAddition (Mitchell and Lapata, 2010) αu + βv 
VAddition au + bv 
Matrix (Socher et al., 2010) g(W[u; v] + b
FullLex (Socher et al., 2012) g(W[Avu;Auv] + b
BiLinear (Socher et al., 2013a,b) g(uEv+W[u;v]+b) 
WMask (Dima, 2015) g(W[uum; vvh] + b
Close Modal

or Create an Account

Close Modal
Close Modal