Our research aims to unify existing works' diverging opinions on how
architectural components affect the adversarial robustness of CNNs. To
accomplish our goal, we synthesize a suite of three generalizable robust
architectural design principles: (a) optimal range for depth and width
configurations, (b) preferring convolutional over patchify stem stage, and (c)
robust residual block design through adopting squeeze and excitation blocks and
non-parametric smooth activation functions. Through extensive experiments
across a wide spectrum of dataset scales, adversarial training methods, model
parameters, and network design spaces, our principles consistently and markedly
improve AutoAttack accuracy: 1-3 percentage points (pp) on CIFAR-10 and
CIFAR-100, and 4-9 pp on ImageNet. The code is publicly available at
this https URL

我们研究的目标是统一现有作品在架构组件如何影响 CNN 的对抗鲁棒性方面的分歧意见。为了实现我们的目标，我们综合了三个可推广的鲁棒架构设计原则：(a) 深度和宽度配置的最佳范围，(b) 优先使用卷积而非拼贴块的初始阶段，和 (c) 通过采用挤压和激励块以及非参数平滑激活函数的鲁棒残差块设计。通过在广泛的数据集规模、对抗训练方法、模型参数和网络设计空间上进行大量实验证明，我们的原则始终显著提高 AutoAttack 的准确性：CIFAR-10 和 CIFAR-100 上为 1-3 个百分点 (pp)，ImageNet 上为 4-9 个百分点 (pp)。代码可以公开访问该 URL。

稳健原则：对抗稳健 CNN 的架构设计原则

Robust Principles: Architectural Design Principles for Adversarially  Robust CNNs

Convolutional neural networks (CNNs) have recently emerged as promising
models of the ventral visual stream, despite their lack of biological
specificity. While current state-of-the-art models of the primary visual cortex
(V1) have surfaced from training with adversarial examples and extensively
augmented data, these models are still unable to explain key neural properties
observed in V1 that arise from biological circuitry. To address this gap, we
systematically incorporated neuroscience-derived architectural components into
CNNs to identify a set of mechanisms and architectures that comprehensively
explain neural activity in V1. We show drastic improvements in model-V1
alignment driven by the integration of architectural components that simulate
center-surround antagonism, local receptive fields, tuned normalization, and
cortical magnification. Upon enhancing task-driven CNNs with a collection of
these specialized components, we uncover models with latent representations
that yield state-of-the-art explanation of V1 neural activity and tuning
properties. Our results highlight an important advancement in the field of
NeuroAI, as we systematically establish a set of architectural components that
contribute to unprecedented explanation of V1. The neuroscience insights that
could be gleaned from increasingly accurate in-silico models of the brain have
the potential to greatly advance the fields of both neuroscience and artificial
intelligence.

我们将神经科学构建的结构组件系统地纳入卷积神经网络，以解释 V1 神经活动和调谐特性，结果证实这些神经科学构建的结构组件可以使 CNN 更好地模拟 V1 神经活动和调谐特性。