Settings for Fgsm() and Pgd().
More...
#include <adversarial.hpp>
|
| double | epsilon = 0.03 |
| | The budget: every adversarial input is within epsilon of its original.
|
| |
| AttackNorm | norm = AttackNorm::Linf |
| |
| int64_t | steps = 10 |
| | PGD: the number of steps.
|
| |
| double | step_size = 0.0 |
| | PGD: the step size; 0 means 2.5 epsilon / steps, Madry et al.'s choice.
|
| |
| bool | random_start = true |
| | PGD: start at a uniformly random point of the ball (the L2 ball for L2), as Madry et al. do; without it PGD starts at the input.
|
| |
| uint64_t | seed = 0 |
| |
| double | clip_min = -std::numeric_limits<double>::infinity() |
| | The valid range of each input element, such as [0, 1] for pixels; every adversarial input is clipped into it. Unbounded by default.
|
| |
| double | clip_max = std::numeric_limits<double>::infinity() |
| |
Settings for Fgsm() and Pgd().
◆ clip_max
| double pulsatrix::AttackOptions::clip_max = std::numeric_limits<double>::infinity() |
◆ clip_min
| double pulsatrix::AttackOptions::clip_min = -std::numeric_limits<double>::infinity() |
The valid range of each input element, such as [0, 1] for pixels; every adversarial input is clipped into it. Unbounded by default.
◆ epsilon
| double pulsatrix::AttackOptions::epsilon = 0.03 |
The budget: every adversarial input is within epsilon of its original.
◆ norm
◆ random_start
| bool pulsatrix::AttackOptions::random_start = true |
PGD: start at a uniformly random point of the ball (the L2 ball for L2), as Madry et al. do; without it PGD starts at the input.
◆ seed
| uint64_t pulsatrix::AttackOptions::seed = 0 |
◆ step_size
| double pulsatrix::AttackOptions::step_size = 0.0 |
PGD: the step size; 0 means 2.5 epsilon / steps, Madry et al.'s choice.
◆ steps
| int64_t pulsatrix::AttackOptions::steps = 10 |
PGD: the number of steps.
The documentation for this struct was generated from the following file: