Thank you for your work,I have recently been reproducing your work. However, I encountered three specific issues during the ablation experiments based on the MFD framework:
1.Hete-loss Performance: When adding only the hete_loss to the MFD baseline, the improvement is not as significant as reported in the paper.
2.Homo-loss Performance: Adding the homo_loss to the MFD baseline actually resulted in a performance drop (negative gain).
3.Missing CT Method Implementation: In the ablation study section, a method referred to as "CT" is mentioned. However, I could not find a corresponding implementation or explicit module named "CT" in the provided source code or paper details.Could you please clarify which part of the source code corresponds to the CT method.
Thank you for your work,I have recently been reproducing your work. However, I encountered three specific issues during the ablation experiments based on the MFD framework:
1.Hete-loss Performance: When adding only the hete_loss to the MFD baseline, the improvement is not as significant as reported in the paper.
2.Homo-loss Performance: Adding the homo_loss to the MFD baseline actually resulted in a performance drop (negative gain).
3.Missing CT Method Implementation: In the ablation study section, a method referred to as "CT" is mentioned. However, I could not find a corresponding implementation or explicit module named "CT" in the provided source code or paper details.Could you please clarify which part of the source code corresponds to the CT method.