AdvRush: Searching for Adversarially Robust Neural Architectures

Mok, Jisoo; Na, Byunggook; Choe, Hyeokjun; Yoon, Sungroh

doi:10.1109/iccv48922.2021.01210

Cited by 14 publications

(11 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Automated network design. As an important branch of AutoML [ 123 ], neural architecture searching (NAS) [ 124 ] has attracted more and more attention. In deep learning-based tasks of classification, detection, segmentation, and tracking, the structure of the neural network has a decisive impact on the performance of the overall algorithm.…”

Section: Discussionmentioning

confidence: 99%

The Synergy between Deep Learning and Organs-on-Chips for High-Throughput Drug Screening: A Review

et al. 2023

View full text Add to dashboard Cite

Organs-on-chips (OoCs) are miniature microfluidic systems that have arguably become a class of advanced in vitro models. Deep learning, as an emerging topic in machine learning, has the ability to extract a hidden statistical relationship from the input data. Recently, these two areas have become integrated to achieve synergy for accelerating drug screening. This review provides a brief description of the basic concepts of deep learning used in OoCs and exemplifies the successful use cases for different types of OoCs. These microfluidic chips are of potential to be assembled as highly potent human-on-chips with complex physiological or pathological functions. Finally, we discuss the future supply with perspectives and potential challenges in terms of combining OoCs and deep learning for image processing and automation designs.

show abstract

Section: Discussionmentioning

confidence: 99%

The Synergy between Deep Learning and Organs-on-Chips for High-Throughput Drug Screening: A Review

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Both GAN and AT are essentially two-player zero-sum games between generator/discriminator and classifier/attacker, respectively which are alternately trained to maximize their respective utilities till convergence corresponding to a Nash Equilibrium (NE) [3]. Mixed-architecture approaches such as MIX+GAN [3], MGAN [32] and DO-GAN in Section 1.2.1 as well as NAS approaches such as Adversarial-NAS [20] and AdvRush [69] have also shown promising results for optimizing the network architecture in AML.…”

Section: Double Oracle Framework For Adversarial Machine Learning (Aml)mentioning

confidence: 99%

“…Moreover, evolutionary algorithm architectures such as R-NAS [45] and RoCo-NAS [21] are restricted to only black-box attacks. Lastly, AdvRush [69] proposes an adversarial robustness-aware neural architecture search algorithm using the input loss landscape of the neural networks to represent their intrinsic robustness.…”

Section: Numerous Approaches Have Been Proposed But Adversarial Train...mentioning

confidence: 99%

“…Moreover, NAS is also used to search the classifier architectures in Adversarial Training (AT) for robustness. AdvRush [69] considers both white-box attacks and black-box attacks using the input loss…”

Section: Chapter Summarymentioning

confidence: 99%

“…Algorithm 6.7 describes the specific changes to adapt DONAS for AT. To obtain the first pair of classifier θ and perturbed dataset δ, we perform neural architecture search for θ by AdvRush [69] and set δ = 0 in InitializeModels(), then store the parameters in the two arrays C and A (line 1-2). For each epoch, we search the new classifier and generate new perturbed dataset with ClassierOracle() and AttackerOracle() using the meta-strategies σ * c and σ * p .…”

Section: Donas For Atmentioning

confidence: 99%

See 2 more Smart Citations