This repo is used for fmars and mumu to cooperate Differential Privacy project.
##Tasks Within this project, tasks will include paper reading, algorithm implementation beside with state-of-the-art technology understanding and creativity proposal.
##Timelines
-
- The project was original created at Feb.2015 and started as a cooperative project at Mar.2015.
-
- We plan to survey the state-of-the-art methods and technologies briefly at Apr.2015 and May.2015.
-
- With high level understanding, we will start to div into specific field at June.2015.
-
- Hope we can work out some new ideas and start to design and implement our prototype at Sept.2015.
-
- If everything goes well, we would like to host mid-term thesis defense at Dec.2015.
-
- Finally accomplish degree paper and graduate requirement at 2016.
##Milestones
##My Innovation point
-
- pruning的时候,采用其他的方式剪枝,已获得更好的决策树——偏向于决策树的改进- -而不是差分隐私。。所以还是需要继续想想看
-
- 昨天的两个会议然后有了一些新的想法:就是是不是现在有些局限了,只看人家的结合方法,很难找到改进的点。所以是不是可以找一些新的决策树分类方法或者其他分类方法如神经网络,考虑将差分隐私应用在这个方法中。这样比较新一些,然后应该创新性也有了。
- 4.23 update:感觉最近确实没啥决策树新方法。。感觉这种算法更新很困难啊。。。你们UCI有米有同学在搞决策树分类啥的?或者其他分类方法的?新一点儿问问看?