Closed facingwaller closed 6 years ago
When you are optimizing top-k performance (instead of the global pairwise ranking performance like AUC) you normally need to focus on the ranking of top k-2k instead of the overall ranking. Thus you need to sample top-ranked negative samples instead of using all the negative samples.
谢谢回答,请问为什么会只需要关注前K-2K的相关的负例?有什么理论依据吗?直觉上来说,不应该是训练了更多的负例,效果更好吗?还是只是把NN模型当做一个黑盒子,从系统的外部上发现了这个规律并不能实证?
You can check out this paper to see why :) http://wnzhang.net/papers/lambdarankcf-sigir.pdf
谢谢回答。我去学习一下这篇论文再向您请教。
请教一下