深度迁移学习技术在金融风控中的应用_老虎社区_美港股上老虎 - 老虎社区

4
评论
收藏

深度迁移学习技术在金融风控中的应用

萨摩耶云科技集团
2021-09-28

文章转载自 DataFunTalk

图 4 长表现期风险任务专家注意力

图 6 多任务gradnorm weighting动态权重对比图

图 7 多任务gradnorm weighting损失函数收敛率对比图

(a) 对抗学习方式的领域适配

(b) 距离度量方式的领域适配图 9 风控场景领域适配建模框架

免责声明：上述内容仅代表发帖人个人观点，不构成本平台的任何投资建议。

4

举报

评论

推荐
最新

empty

暂无评论

热议股票

关于我们·老虎社区守则·老虎社区账号管理规范·老虎社区服务协议·老虎社区隐私政策

公司名称：北京至简风宜信息技术有限公司

违法和不良信息投诉：010-5681-3562（工作时间9:30-18:30）

邮箱：service@laohu8.com

老虎社区侵权投诉指引

中国互联网举报中心·北京互联网举报中心

网络谣言曝光台·扫黄打非举报中心

网络110报警服务

© 2018-2024 老虎社区版权所有

营业执照：91110105MA01A4U55R

ICP备：京ICP备18016422号

网安备：京公网安备11010502036490

qrcode

关注我们

{"i18n":{"language":"zh_CN"},"data":{"magic":2,"id":862053559,"tweetId":"862053559","gmtCreate":1632820600668,"gmtModify":1632820600668,"author":{"id":4094974360355080,"idStr":"4094974360355080","authorId":4094974360355080,"authorIdStr":"4094974360355080","name":"萨摩耶云科技集团","avatar":"https://static.tigerbbs.com/1ff074b1915aa597410906b83515e00a","vip":5,"userType":5,"introduction":"中国领先的AI决策解决方案提供商","boolIsFan":false,"boolIsHead":false,"crmLevel":1,"crmLevelSwitch":0,"individualDisplayBadges":[],"fanSize":1429,"starInvestorFlag":false},"themes":[],"images":[{"img":"https://static.tigerbbs.com/2e763447efc52c93dc9bcd2addbb5ade","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/dd81391452e7e52c0b2312efcc9280a5","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/c42d72e1685add6418abd99effae6782","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/9100d84fdfa179b610d9ca4e46534319","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/c058b0741cc487dc53bd786d6e93bbaf","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/b07cdfb09f2f59853f934313a5a4184e","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/eb28adef240586ebf15d60287740e626","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/cb6e39b11dea1cfb72cfc1f8cd474077","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/724f226827b2c6f96a06b50edb10949e","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/5bdbd9621ec089f9f245ac6b8e871b9a","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/e1404b80c260e166b3b9bc68630a5f7a","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/fa0e6b90362c05062a2dc99e1faa707c","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/d29b7b534b48e6fb087d989b9ddca80d","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/c5578b51d434c9cf9e62fa3224d73911","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/29561ed7cc89a9f69b8941d891375178","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/6749bb613fd8a45575015db47ccad71d","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/7e31eecd7921b0d6b3ecd7f9e0b027f3","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/f894cd46700732467badc7e79c84eed2","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/c8a9f3f2c7343ad560b7d90ff5a3d301","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/a4630c62d92a1b04b1514c2f1dbca652","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/fabffc98cdc4d46ac93d178969787801","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/115a97c00875ca598b804c92e0d12e81","width":"-1","height":"-1"}],"coverImages":[{"img":"https://static.tigerbbs.com/2e763447efc52c93dc9bcd2addbb5ade","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/dd81391452e7e52c0b2312efcc9280a5","width":"-1","height":"-1"},{"img":"https://static.tigerbbs.com/c42d72e1685add6418abd99effae6782","width":"-1","height":"-1"}],"title":"深度迁移学习技术在金融风控中的应用","extraTitle":"","html":"<html><head></head><body><p><i>文章转载自 DataFunTalk</i></p> <p><img src=\"https://static.tigerbbs.com/fabffc98cdc4d46ac93d178969787801\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/b07cdfb09f2f59853f934313a5a4184e\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/cb6e39b11dea1cfb72cfc1f8cd474077\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/115a97c00875ca598b804c92e0d12e81\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/29561ed7cc89a9f69b8941d891375178\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/6749bb613fd8a45575015db47ccad71d\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/724f226827b2c6f96a06b50edb10949e\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/c8a9f3f2c7343ad560b7d90ff5a3d301\" tg-width=\"-1\" tg-height=\"-1\"><span>图 4 长表现期风险任务专家注意力</span></p> <p><img src=\"https://static.tigerbbs.com/e1404b80c260e166b3b9bc68630a5f7a\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/a4630c62d92a1b04b1514c2f1dbca652\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/f894cd46700732467badc7e79c84eed2\" tg-width=\"-1\" tg-height=\"-1\"><span>图 6 多任务gradnorm weighting动态权重对比图</span></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/c058b0741cc487dc53bd786d6e93bbaf\" tg-width=\"-1\" tg-height=\"-1\"><span>图 7 多任务gradnorm weighting损失函数收敛率对比图</span></p> <p><img src=\"https://static.tigerbbs.com/2e763447efc52c93dc9bcd2addbb5ade\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/9100d84fdfa179b610d9ca4e46534319\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/fa0e6b90362c05062a2dc99e1faa707c\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/eb28adef240586ebf15d60287740e626\" tg-width=\"-1\" tg-height=\"-1\"><span>(a) 对抗学习方式的领域适配</span></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/dd81391452e7e52c0b2312efcc9280a5\" tg-width=\"-1\" tg-height=\"-1\"><span>(b) 距离度量方式的领域适配 图 9 风控场景领域适配建模框架</span></p> <p><img src=\"https://static.tigerbbs.com/d29b7b534b48e6fb087d989b9ddca80d\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/c5578b51d434c9cf9e62fa3224d73911\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/c42d72e1685add6418abd99effae6782\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/5bdbd9621ec089f9f245ac6b8e871b9a\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/7e31eecd7921b0d6b3ecd7f9e0b027f3\" tg-width=\"-1\" tg-height=\"-1\"></p></body></html>","htmlText":"<html><head></head><body><p><i>文章转载自 DataFunTalk</i></p> <p><img src=\"https://static.tigerbbs.com/fabffc98cdc4d46ac93d178969787801\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/b07cdfb09f2f59853f934313a5a4184e\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/cb6e39b11dea1cfb72cfc1f8cd474077\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/115a97c00875ca598b804c92e0d12e81\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/29561ed7cc89a9f69b8941d891375178\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/6749bb613fd8a45575015db47ccad71d\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/724f226827b2c6f96a06b50edb10949e\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/c8a9f3f2c7343ad560b7d90ff5a3d301\" tg-width=\"-1\" tg-height=\"-1\"><span>图 4 长表现期风险任务专家注意力</span></p> <p><img src=\"https://static.tigerbbs.com/e1404b80c260e166b3b9bc68630a5f7a\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/a4630c62d92a1b04b1514c2f1dbca652\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/f894cd46700732467badc7e79c84eed2\" tg-width=\"-1\" tg-height=\"-1\"><span>图 6 多任务gradnorm weighting动态权重对比图</span></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/c058b0741cc487dc53bd786d6e93bbaf\" tg-width=\"-1\" tg-height=\"-1\"><span>图 7 多任务gradnorm weighting损失函数收敛率对比图</span></p> <p><img src=\"https://static.tigerbbs.com/2e763447efc52c93dc9bcd2addbb5ade\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/9100d84fdfa179b610d9ca4e46534319\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/fa0e6b90362c05062a2dc99e1faa707c\" tg-width=\"-1\" tg-height=\"-1\"></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/eb28adef240586ebf15d60287740e626\" tg-width=\"-1\" tg-height=\"-1\"><span>(a) 对抗学习方式的领域适配</span></p> <p class=\"t-img-caption\"><img src=\"https://static.tigerbbs.com/dd81391452e7e52c0b2312efcc9280a5\" tg-width=\"-1\" tg-height=\"-1\"><span>(b) 距离度量方式的领域适配 图 9 风控场景领域适配建模框架</span></p> <p><img src=\"https://static.tigerbbs.com/d29b7b534b48e6fb087d989b9ddca80d\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/c5578b51d434c9cf9e62fa3224d73911\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/c42d72e1685add6418abd99effae6782\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/5bdbd9621ec089f9f245ac6b8e871b9a\" tg-width=\"-1\" tg-height=\"-1\"><img src=\"https://static.tigerbbs.com/7e31eecd7921b0d6b3ecd7f9e0b027f3\" tg-width=\"-1\" tg-height=\"-1\"></p></body></html>","text":"文章转载自 DataFunTalk 图 4 长表现期风险任务专家注意力 图 6 多任务gradnorm weighting动态权重对比图 图 7 多任务gradnorm weighting损失函数收敛率对比图 (a) 对抗学习方式的领域适配 (b) 距离度量方式的领域适配 图 9 风控场景领域适配建模框架","highlighted":1,"essential":1,"paper":2,"likeSize":4,"commentSize":0,"repostSize":4,"favoriteSize":0,"link":"https://laohu8.com/post/862053559","repostId":0,"isVote":1,"tweetType":1,"viewCount":11182,"commentLimit":10,"likeStatus":false,"favoriteStatus":false,"reportStatus":false,"symbols":[],"verified":2,"subType":0,"readableState":1,"langContent":"CN","currentLanguage":"CN","warmUpFlag":false,"orderFlag":false,"shareable":true,"causeOfNotShareable":"","featuresForAnalytics":[],"commentAndTweetFlag":false,"andRepostAutoSelectedFlag":false,"upFlag":false,"length":213,"xxTargetLangEnum":"ZH_CN"},"commentList":[],"hasMoreComment":false,"orderType":2}