how to feed and care for your machine-learning experts

how to feed and care for your machine-learning experts

ID:7279650

大小:497.12 KB

页数:10页

时间:2018-02-10

how to feed and care for your machine-learning experts_第1页
how to feed and care for your machine-learning experts_第2页
how to feed and care for your machine-learning experts_第3页
how to feed and care for your machine-learning experts_第4页
how to feed and care for your machine-learning experts_第5页
资源描述:

《how to feed and care for your machine-learning experts》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库

1、CHAPTER16HowtoFeedandCareforYourMachine-LearningExpertsPeteWardenMachinelearningisacraftaswellasascience,andforthebestresultsyou’lloftenneedtoturntoexperiencedspecialists.Noteveryteamhasenoughinterestingproblemstojustifyafull-timemachine-learningposition,

2、though.Asthevalueoftheapproachbe‐comesbetter-known,thedemandforpart-timeorproject-basedmachine-learningworkhasgrown,butit’softenhardforatraditionalengineeringteamtoeffectivelyworkwithoutsideexpertsinthefield.I’mgoingtotalkaboutsomeofthethingsIlearnedwhile

3、runninganoutsourcedprojectthroughKaggle,1acommunityofthousandsofresearcherswhoparticipateindatacompetitionsmodeledontheNetflixPrize.Thiswasanextremeexampleofoutsourcing:weliterallyhandedoveradataset,ashortdescrip‐tion,andasuccessmetrictoalargegroupofstran

4、gers.Ithadalmostnoneofthetra‐ditionalinteractionsyou’dexpect,butitdidteachmevaluablelessonsthatapplytoanyinteractionswithmachine-learningspecialists.DefinetheProblemMycompanyJetpaccreatesatravelmagazinewrittenbyyourfriends,usingvacationphotosthey’veshared

5、withyouonFacebookandothersocialservices.Theaverageuserhashadovertwohundredthousandpicturessharedwiththem,sowehavealottochoosefrom.Unfortunately,manyofthemarenotverygood,atleastforourpurposes.Whenweshowedpeopleourprototype,mostwouldbeturnedoffbythepoorqual

6、ityoftheimages,soweknewweneededasolutionthatwouldhelpuspickoutthemostbeautifulfromthedeeppooltowhichwehadaccess.1.“Kaggle:makingdatascienceasport.”(http://www.kaggle.com/)195Thisintentionwastoovaguetobuildon,though.Wehadtodecideexactlywhatwewantedourproce

7、sstoproduce,andhencewhatquestionwewantedtoanswerabouteachphoto.Thisisthecrucialfirststepforanymachine-learningapplication.Unlessyouknowwhatyourgoalis,you’llneverbeabletobuildasuccessfulsolution.Oneofthekeyqualitiesyouneedissomedegreeofobjectivityandrepeat

8、ability.If,likeme,you’retryingtopredictahuman’sreaction,it’snogoodifitvariesunpredictablyfrompersontoperson.Asking“Isthisagoodflavoroficecream?”mightbeagoodmarketresearchquestion,buttheresultsaysmoreaboutanindividua

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。