easezyc / metaheac Goto Github PK

您好，谢谢这么详尽的code implementation！想请问下，如果一个batch里同时有正例和负例，在取user embedding的mean的时候是不是把负例也算进去了呀？还是说user embedding的mean 是固定的，是从campaign level计算得到的？谢谢！

您好：
腾讯Look-alike数据集和预处理后的数据集的链接都已经失效了，能否提供新的有效链接？
万分感谢！

您好，请问这个Wechat数据集是不是不方便提供的？

Hello, sorry to disturb you. I want to ask if you have compared the experiment result with 2018's submit results? Because I found that many participants reported the auc aboce 0.75, which is much higher than the aucs listed in the paper. Why?

there is a mistake in python main.py --task_count 5 --num_expert 8 --output 5 --batchsize 512

output is not a right args which should be modified to num_output