Multiple Adversarial Domains Adaptation Approach for Mitigating Adversarial Attacks Effects

Algorithm 1

MADA algorithm.

	Require: A clean dataset , minibatch size , learning rates , critic training step , adversarial domains d
1	Initialize feature generator , domain critic , and classifier , with weights
2	Repeat
3	Sample clean minibatch
4	for do
5	Use the current state of to generate adversarial examples and create adversarial minibatch
6	end for
7	while do
8	Find ,
9
10	end while
11	Find ,
12
13