什么是“安全变量”？
我刚刚读了一篇关于不同排名算法的文章。我有点困惑的一件事是 stumbleupon 的算法：
（初始绊脚石受众/# 域） + ((% 绊倒者观众 / # 个域名) + 有机奖励 – 非朋友) – (% 绊脚石观众 + 有机奖金）+ N
N 是一个“安全变量”，因此 假设算法是灵活的。它 代表一个随机数。
有人可以向我解释一下 N 的用途吗？我不明白它如何使算法更加灵活。
I was just reading an article about different ranking algorithms. One thing I'm a bit confused about is stumbleupon's algoritm:
(Initial stumbler audience / # domain)
+ ((% stumbler audience / # domain) + organic bonus – nonfriend) – (%
stumbler audience + organic bonus) + NN is a "safety variable" so that the
assumed algorithm is flexible. It
represents a random number.
Could someone explain the purpose of N to me? I don't understand how it's making the algorithm more flexible.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
它会导致结果以不可预测的速度下降。如果你曾经跌跌撞撞地跌跌撞撞了几个小时，那么当你开始摸索时，你可能会注意到每次跌跌撞撞的质量和相关性都严重下降。
如果没有随机变量，结果将以非常可预测的速度下降。您最可能喜欢的链接将是第一个，而后续的每个链接对您的吸引力可能会稍差一些。该变量试图在某种程度上否定这一点。
这是 stumbleupon 旨在让人上瘾的众多巧妙方法之一：
It helps the results degrade at an unpredictable rate. If you've ever stumbled for a couple of hours you'll likely have noticed a serious decline in the quality and relevance of each stumble as you start scraping the bottom of the barrel.
Without the random variable the results would degrade in a very predictable rate. The link you were most likely to love would be first and each subsequent link would likely appeal a little less to you. The variable attempts to negate this somewhat.
It's one of the many clever ways stumbleupon is designed to be addictive:
通过引入 N，您可以确保那些不经常被评级的内容偶尔会被洗牌到顶部（如果 N 结果很高）。我想这可以确保所有东西最初至少被看到几次，从而有机会获得更多的绊脚石（或者无论它们被称为什么）。
基本上只是对新内容进行轻微的随机化，以消除排名第一的偏见。
无论如何，这是我的猜测。
By introducing N you ensure that stuff that is not rated very often yet gets shuffled to the top occassionally (if N turns out high). I guess this ensures that everything gets seen at least a few times initially, and thus gets a few shots at obtaining more stumbles (or whatever they're called).
Basically just slight randomization of new stuff to remove top-spot bias.
That is my guess anyway.