LTR Loss issue

[graph-0]

LTR Loss issue

[GRAPH-0]

Thanks for your interest in this paper.
The LTR loss can be regarded as an auxiliary loss function of the RMSE loss.
Ideally, the RMSE loss pushes the predicted value close to the ground truth, and the LTR loss is more concerned with the relative numerical size between the entity importance scores. In the experiment, there are related hyperparameters that control the weight of these two terms. In this way, we better model the relative relationship of importance between entities.

[Aurora-LiZhi]

Thanks for your interest in this paper. The LTR loss can be regarded as an auxiliary loss function of the RMSE loss. Ideally, the RMSE loss pushes the predicted value close to the ground truth, and the LTR loss is more concerned with the relative numerical size between the entity importance scores. In the experiment, there are related hyperparameters that control the weight of these two terms. In this way, we better model the relative relationship of importance between entities.

Ok , I konw it. Thank you