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.

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.

Abstract: The present invention is related to a filtering device with slotted ground structure, comprising a first substrate, a second substrate and a pair of differential signal lines, in which a ground plane having slotted ground structure is provided between the first substrate and the second substrate. Each of the two differential signal lines is symmetric to each other and comprises a first line segment being horizontally provided on the top surface of the first substrate and a second line segment being horizontally provided on the bottom surface of the second substrate, respectively. The first line segment is connected to the second line segment through a vertically disposed conductive via. Thereby, a common-mode noise within a specific frequency band may be suppressed effectively, so as to avoid interference a differential-mode signal transmitted on the differential signal lines, due to the slotted ground structure etched on the ground plane.