Hyperparameters for both Conv-TransE and SACN models

JD-AI-Research-Silicon-Valley / SACN

End-to-end Structure-Aware Convolutional Networks for Knowledge Base Completion

MIT License

112 stars 31 forks source link

Hyperparameters for both Conv-TransE and SACN models #8

Open Nilesh-Agrawal opened 5 years ago

Nilesh-Agrawal commented 5 years ago

Can you please share the hyperparameters you used to reproduce the results mentioned in paper.

chaoshangcs commented 5 years ago

Thanks for your message. Please check the answered issue. Others have reproduced the results. First, please run the latest version of code with more epochs (such as 1000 epochs). Second, The model is sensitive to the parameters. Since we have reorganized the code, some hyper-parameters have changed a little bit. Tuning the init_emb_size, dropout_rate, channels and lr parameters will help to improve the performance. If you have more questions, please email me. Thanks!