README.md

TransH

Reimplement TransH mode

file

TransH.py _The initial version which use my own code to implement.
TransH_pytorch.py _Try to use the autograd function of pytorch to calculate the gradient of each embedding.
TransH_torch.py _Use the torch.nn.Embedding and marginRankingLoss function to reimplement the TransH model.

TransH_torch.py 中完全使用pytorch实现transH模型，以及测试功能，在FB15K 以及 WN18两个数据集上的Link prediction测试结果效果基本达到论文原文中的测试结果

TransH_torch.py Test_result

WN18:

	raw	fil
	MR hit10	MR hit10
TransH	253 0.6719	242 0.768

FB15K:

	raw	fil
	MR hit10	MR hit10
TransH	327 0.457	188 0.663