Thanks for your excellent work, but when I switch the model to transE while fixing other parameters unchanged, the top@10 retrieval rate doesn't improve during training and is still below to 1% after training for 300 epochs. The result looks good for ComplEx. Here is my command for training:
The hyper-parameters suitable for ComplEx may not be suitable for TransE. You can try to adjust the hyper-parameters, such as learning rate, weight decay, loss function, etc., first.
Thanks for your excellent work, but when I switch the model to transE while fixing other parameters unchanged, the top@10 retrieval rate doesn't improve during training and is still below to 1% after training for 300 epochs. The result looks good for ComplEx. Here is my command for training:
git clone https://github.com/yzhangee/NSCaching cd NSCaching git clone https://github.com/thunlp/OpenKE mv OpenKE/benchmarks ../KG_Data python train.py --model TransE