search

easezyc / Multitask-Recommendation-Library

MTReclib provides a PyTorch implementation of multi-task recommendation models and common datasets.
MIT License
282 stars 45 forks source link

请问为什么表格中 第二个任务是 “CTCVR” 而不是 “CVR” #12

Open 2catycm opened 1 year ago

2catycm commented 1 year ago

按照理解，第二个表格评估的时候，是直接用CVR任务的y_true 和 模型的预测 y_pred 去评估的。 如果是CTCVR的话，评估的时候，应该首先减小数据集，把数据集筛选到只剩下 CTR=1的情况下，然后此时CVR这个label才变成了CTCVR。而相应的模型预测概率，也应该是在看到第一个标签的情况下训练的模型所预测出来的概率，才叫做CTCVR

furkandurmus commented 4 months ago

There can't be any sample like CTR:0 CVR:1 if CVR is 1 that means its CTR is already 1 so I don't think we need to filter out CTR:1 samples. Eventually CVR and CTCVR are just same here isn't it?