alibaba / alibaba/x-deeplearning
ESMM数据负采样预估修正问题
Open
- Dominant language
- PureBasic
- Stars
- 4.3k
- Forks
- 1k
- PR merge metrics
- No merged PRs in 30d
Description
请问一下各位大佬,有两个问题想请教一下:
1.
ESMM做曝光负采样后,模型训练得到的pctr,pctcvr都会偏高,是否需要在模型训练时进行修正,还是需要在训练结束后修正?
2.
在实际应用中,是直接预测pctcvr来作为排序指标,还是使用ctr*cvr的方式?ctr可以是别的模型生成,cvr由esmm生成,谢谢
Contributor guide
No contributing guide indexed for this repository
Research direction
The issue mentions ESMM, exposure negative sampling, pCTR, and pCVR but names no files, tests, or entry points. First clarify whether a code change is requested; completion would require an agreed answer about when to correct the estimates and whether ranking should use pCVR or CTR×CVR.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- machine-learning
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100