<ul><li>Transfer-based attacks generate adversarial examples on the surrogate model, which can mislead other black-box models.</li><li>Non-linear layers like ReLU and max-pooling truncate the gradient during backward propagation, undermining the transferability of adversarial examples.</li><li>A novel method called Backward Propagation Attack (BPA) is proposed to increase the relevance between the gradient and loss function, boosting adversarial transferability.</li><li>Conditional Backward Propagation of Chaos investigates the well-posedness of a backward stochastic differential equation and proves propagation of chaos in an associated interacting particles system.</li></ul>

Backward Propagation is here to stay part2(AI 2024)

Discover more