However, SSA T suffers from catastrophic overfit-ting (CO), a phenomenon that leads to a severely distorted classifier, making it vulnerable to multi-step adversarial attacks. In this work, we observe that some adversarial examples generated on the SSA T -trained network exhibit anomalous behaviour, that is, although these training samples are generated by the inner maximization process, their associated loss decreases instead, which we named abnormal adversarial examples (AAEs).

Neural Information Processing Systems http://nips.cc/

Fortheformer setting, an extensive literature has explored the statistical and computational aspects of learning Booleanfunctions[Angluin,1992,HellersteinandServedio,2007].

A recent trend in online algorithms has seen the use of future predictions generated by ML techniques to bypass pessimistic worst-case lower bounds.

Wealsoshow thatacertified version ofthecomputationally tractable DOO algorithm matches these packing and integral bounds.

Then, we propose a series of DP-SignRP algorithms that leverage the robustness of the "sign flipping probability" of random projections.