亚博app

计较机迷信与工程学院系列学术报告——李罡 澳大利亚迪肯大学

作者:常婉纶 考核人: 点击:[]

baogaobiaotiwenti:privacy aware data science

baogaoshihou:2020nian12yue4ri zhouwu xiazhanshu2:30

tengxunjihuiid: 979 496 572

baogaoren:ligang boshi aodaliyadikendaxuexinxishouyixueyuan

zhuanjiajianjie:gang li,nan,aodaliyadikendaxuexinxishouyixueyuanyantaoyuan,aodaliyadikendaxuedata to intelligenceshujuzhinengyantaozhongjianzhuren。

jinchaodanfuieeejijiaozhinengxuehuishujufajueyudashujuchanfadmtcshouyiweiyuanhuiweiyuan(2017-2018fuzhuxi)、ieee smcxuehuichanyexinxitixi(tceis)shouyiweiyuanhuiweiyuan、ieee task force on educational data miningzhuxi (2020-2023),bingqiedanfusciqikanjournal of travel research (sage), decision support systems(elsevier)、ieee access (ieee)he information discovery & delivery (emerald)fubianjihezazhibianwei。2017/2020changshimixingongchengyubanliguojijihui(ksem2017/2020,ccf c)fashizhuxihejihuizhuxi、2016 ieeechanyexinxitixijihui(es2016)fashizhuxi、2016zhinengxinxichuzhijihui(iip2016)fashizhuxi、2016 bescjihuifashizhuxi、2019changshimixingongchengyubanliguojijihui(ksem2019,ccf c)dahuizhuxi 。

baogaozeyao:this talk will present an overview of recent advances in differential privacy, in the area of data science. after recapping the basic concepts on differential privacy, we will introduce the requirements of differential privacy in different tasks of data science, including pac learning, objective optimization and unsupervised learning. then, techniques for bringing differential privacy into those tasks will be categorized as input perturbation, objective perturbation and output perturbation. a strong composition theorem to quantify the privacy loss in the context of data science will be introduced, as an extension of the simple composition and the advance composition theorems.

下一条:陕西师范大学魏俊发、曹睿传授特邀报告

封闭

saoyisaofenxiangbenye

亚博app 亚博网页版 亚博网页版 亚博网页版 爱游戏app