We present an adversarial framework to craft perturbations that mislead classifiers by accounting for the image content and semantics of labels. The proposed combines a structure loss semantic in multi-task objective function train fully convolutional neural network. helps generate whose type magnitude are defined target processing filter. considers groups (semantic) labels prevent filtered {fr...

If the originally flat bottom of a wide quantum well with multiple eigenstates is periodically modulated, its eigenvalues rearrange into denser groups separated by wider gaps. We show that this effect, if implemented in an elongated bottle microresonator [also called surface nanoscale axial photonics (SNAP) microresonator] allows us to design microwave photonic tunable filters outstanding perfo...

in this paper a particular case of z-ideals, called strongly z-ideal, is defined by introducing zero sets in pointfree topology. we study strongly z-ideals, their relation with z-ideals and the role of spatiality in this relation. for strongly z-ideals, we analyze prime ideals using the concept of zero sets. moreover, it is proven that the intersection of all zero sets of a prime ideal of c(l),...