Invention Grant
- Patent Title: Framework for certifying a lower bound on a robustness level of convolutional neural networks
-
Application No.: US16256267Application Date: 2019-01-24
-
Publication No.: US11625487B2Publication Date: 2023-04-11
- Inventor: Pin-Yu Chen , Sijia Liu , Akhilan Boopathy , Tsui-Wei Weng , Luca Daniel
- Applicant: International Business Machines Corporation , MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- Applicant Address: US NY Armonk; US MA Cambridge
- Assignee: International Business Machines Corporation,MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- Current Assignee: International Business Machines Corporation,MASSACHUSETTS INSTITUTE OF TECHNOLOGY
- Current Assignee Address: US NY Armonk; US MA Cambridge
- Agency: McGinn I.P. Law Group, PLLC.
- Agent Peter Edwards, Esq.
- Main IPC: G06F21/57
- IPC: G06F21/57 ; G06N3/08 ; G06N20/00 ; G06N3/04
Abstract:
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.
Public/Granted literature
- US20200242252A1 FRAMEWORK FOR CERTIFYING A LOWER BOUND ON A ROBUSTNESS LEVEL OF CONVOLUTIONAL NEURAL NETWORKS Public/Granted day:2020-07-30
Information query
