CNN METHOD AND DEVICE FOR VERIFYING INTEGRITY OF PARAMETERS OF CNN BY USING TEST PATTERN TO ENHANCE FAULT TOLERANCE AND FLUCTUATION ROBUSTNESS IN EXTREME SITUATIONS FOR FUNCTIONAL SAFETY

摘要

The present invention provides at least one original input in order to improve fault tolerance and flux robustness and reduce annotation costs in extreme situations for functional safety on a convolutional neural network (CNN). It relates to a method for verifying the integrity of at least one parameter of a CNN using at least one test pattern added to, (a) the computing device, the at least one Adding Unit (Adding Unit), the original test pattern input Generating at least one extended input by adding to; (b) causing the computing device to generate at least one output for verification by causing the CNN to apply a convolution operation at least once to the extended input; And (c) allowing the computing device to verify the integrity of the parameters of the CNN by determining the validity of the verification output by referring to the at least one comparison unit. It provides a method comprising a; step.