- 专利标题： Framework for certifying a lower bound on a robustness level of convolutional neural networks
-
申请号： US16256267申请日： 2019-01-24
-
公开(公告)号： US11625487B2公开(公告)日： 2023-04-11
- 发明人: Pin-Yu Chen , Sijia Liu , Akhilan Boopathy , Tsui-Wei Weng , Luca Daniel
- 申请人： International Business Machines Corporation , MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- 申请人地址： US NY Armonk; US MA Cambridge
- 专利权人： International Business Machines Corporation,MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- 当前专利权人： International Business Machines Corporation,MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- 当前专利权人地址： US NY Armonk; US MA Cambridge
- 代理机构： McGinn I.P. Law Group, PLLC.
- 代理商 Peter Edwards, Esq.
- 主分类号： G06F21/57
- IPC分类号： G06F21/57 ; G06N3/08 ; G06N20/00 ; G06N3/04
摘要：
A certification method, system, and computer program product include certifying an adversarial robustness of a convolutional neural network by deriving an analytic solution for a neural network output using an efficient upper bound and an efficient lower bound on an activation function and applying the analytic solution in computing a certified robustness.
公开/授权文献
信息查询
