DARTSRepair: Core-failure-set guided DARTS for network robustness to common corruptions. (November 2022)

• Record Type: Journal Article
• Authors: Ren, Xuhong; Chen, Jianlang; Juefei-Xu, Felix; Xue, Wanli; Guo, Qing; Ma, Lei; Zhao, Jianjun; Chen, Shengyong
• Is Part Of: Pattern recognition. Volume 131(2022)
• Publication Date: 2022-11
• Subjects: Network architecture search -- Core-failure-set selection -- Robustness enhancement -- Differentiable architecture search; 07.05.Mh; 68T10; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108864
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We focus on an important but challenging problem that is rarely explored, that is, how to enhance the robustness of deployed deep neural network (DNN) via the guidance of a few collected and misclassified examples that might containing unknown corruptions. We first conduct an empirical study and validate that the model architectures can be definitely related to the corruptions having a specific pattern. We propose a novel core-failure-set guided DARTS that embeds a Kcenter-greedy algorithm for DARTS to select suitable corrupted failure examples to refine the model architecture. Compared with the state-of-the-art NAS and data-augmentation-based enhancement methods, our final method can achieve higher accuracy on all corrupted datasets and the original clean dataset. In particular, on some of the corruptions, we can achieve over 45% absolute accuracy improvements. Abstract: Network architecture search (NAS), in particular the differentiable architecture search (DARTS) method, has shown a great power to learn excellent model architectures on the specific dataset of interest. In contrast to using a fixed dataset, in this work, we focus on a different but important scenario for NAS: how to refine a deployed network's model architecture to enhance its robustness with the guidance of a few collected and misclassified examples that are degraded by some real-world unknown corruptions having a specific pattern ( e.g ., noise, blur, etc ..). To this end, we first conduct an empirical study to validate that the model architectures can be definitely related to the corruption patterns. Surprisingly, by just adding a few corrupted and misclassified examples ( e.g ., 10 3 examples) to the clean training dataset ( e.g ., 5.0 × 10 4 examples), we can refine the model architecture and enhance the robustness significantly. To make it more practical, the key problem, i.e ., how to select the proper failure examples for the effective NAS guidance, should be carefully investigated. Then, we propose a novel core-failure-set guided DARTS that embeds a K -center-greedy algorithm for DARTS to select suitable corrupted failure examples to refine the model architecture. We use our method for DARTS-refined DNNs on the clean as well as 15 corruptions with the guidance of four specific real-world corruptions. Compared with the state-of-the-art NAS as well as data-augmentation-based enhancement methods, our final method can achieve higher accuracy on both corrupted datasets and the original clean dataset. On some of the corruption patterns, we can achieve as high as over 45 % absolute accuracy improvements.