EURASIP Journal on Image and Video Processing

Table 2 Neural embedding scoring and loss functions

From: Semantic embeddings of generic objects for zero-shot learning

Model	ψ(e_s,r_p,e_o)	\(\mathcal {L}\)
TransE [26]	∥e_s+r_p−e_o∥	L ₂
DistMult [25]	〈e_s,r_p,e_o〉	Ranking
ConvE [27]	f(vec(f([e_s;r_p]∗w))W)e_o	BCE
TransE^∗	〈e_s+r_p,e_o〉	Triplet

Back to article page