arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2303.12669cs.CVcs.AIcs.LG
收藏

对抗训练下类人行为的扩展研究

An Extended Study of Human-like Behavior under Adversarial Training

Offenburg University(奥芬堡大学);Fraunhofer ITWM(弗劳恩霍夫工业数学研究所);University of Siegen(锡根大学);Max Planck Institute for Informatics(马克斯·普朗克信息学研究所)

Paul Gavrikov, Janis Keuper, Margret Keuper

摘要:

神经网络有许多缺点。其中最重要的是对分布偏移的敏感性,这使得模型很容易被对人类通常无法感知的输入的小扰动所欺骗,并且不必携带语义含义。对抗训练通过在最坏情况下的扰动上训练模型来解决这个问题。然而,最近的研究也指出,神经网络的推理与人类不同。人类通过形状识别物体,而神经网络主要使用纹理线索。例如,在照片上训练的模型可能无法推广到包含草图的数据集。有趣的是,研究还表明,对抗性训练似乎有利于增加向形状偏见的转变。在这项工作中,我们重新审视了这一观察,并提供了广泛的分析,这种影响的各种架构,常见的$\ell_2$和$\ell_\infty$的培训,和基于transformer的模型。此外,我们从频率的角度为这种现象提供了一种可能的解释。

英文摘要:

Neural networks have a number of shortcomings. Amongst the severest ones is the sensitivity to distribution shifts which allows models to be easily fooled into wrong predictions by small perturbations to inputs that are often imperceivable to humans and do not have to carry semantic meaning. Adversarial training poses a partial solution to address this issue by training models on worst-case perturbations. Yet, recent work has also pointed out that the reasoning in neural networks is different from humans. Humans identify objects by shape, while neural nets mainly employ texture cues. Exemplarily, a model trained on photographs will likely fail to generalize to datasets containing sketches. Interestingly, it was also shown that adversarial training seems to favorably increase the shift toward shape bias. In this work, we revisit this observation and provide an extensive analysis of this effect on various architectures, the common $\ell_2$- and $\ell_\infty$-training, and Transformer-based models. Further, we provide a possible explanation for this phenomenon from a frequency perspective.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑