阅读众包文献中一些值得mark 的小收获

1. Power Law distribution

来自 Whom to Ask? Jury Selection for Decision Making Tasks on Micro-blog Serves

和 Community-Based Bayesian Aggregation 和 Aggregating Crowdsourced Binary Ratings

 

 

2. Anchoring effect

来自 Whom to Ask? Jury Selection for Decision Making Tasks on Micro-blog Serves

还在上一篇文章中见过，但不记得了，后面记起来了再补充。

 

3. uninformative priors

来自 Sequential crowdsourced labeling as an epsilon-greedy exploration in a Markov Decision Process

19年寒假重读PRML 前两章时发现也讲了此

 

4. machine-learning based vs. linear-algebraic based

machine-learning based 通常依赖于 EM 算法，其对工人-任务分配图上没有要求，但不提供任何最终结果的理论保证。

linear-algebraic based 通常需要任务分配图是 random regular 或者是 complete, 这样才可以提供理论保证。

通常众包中的算法可以分为这两类，这一说法最早来自于 paper Aggregating Crowdsourced Binary Ratings

在后一篇paper Reputation-based worker Filtering in Crowdsourcing 也用到的这一说法。

 

4. Expectation-Propagation (EP) 消息传递算法

community-Based Bayesian Aggragtion models for Crowdsourcing

应该 PRML 上也讲了此