Research Article

Exploring Security Vulnerabilities of Deep Learning Models by Adversarial Attacks

Table 8

A generated adversarial example by the VSWR algorithm (Amazon review dataset).

Original textAdversarial example

Textwhile not expensive to initially purchase, this system is expensive in the long run. if you mop your floors regularly or have a lot of rooms to mop, you'll be constantly replacing pads and cleaner and batteries, none of which are cheap.additionally, this system does not clean my laminate floors well. it leaves them cloudy/hazy and they never look or feel 100% clean afterwards. when i go over with a damp paper towel where i just mopped, it always shows a fair amount of dirt. and looking at the pad when you remove it from the mop you can see why. the way it is designed, the most pressure is on the front edge of the mop. the center and back of the pad do not receive nearly as much pressure and pick up almost nothing.in short, for small jobs and irregular cleaning, this mop will do. otherwise, it is inefficient, expensive and ultimately frustrating.while not expensive to initially purchase, this system is expensive in the long run. if you mop your floors regularly or have a lot of rooms to mop, you'll be constantly replacing pads and cleaner and batteries, none of which are cheap.additionally, this system does not clean my laminate floors well. it leaves them cloudy/hazy and they never look or feel 100% clean afterwards. when i go over with a damp paper towel where i just mopped, it always shows a fair amount of dirt. and looking at the pad when you remove it from the mop you can see why. the way it is designed, the most pressure is on the front edge of the mop. the center and back of the pad do not receive nearly as much pressure and pick up almost nothing.in short, for small jobs and irregular cleaning, this mop will do. otherwise, it is inefficient, expensive andd ultimately frustarating.
LabelNegativePositive