Abstract:Over the past few years, deep neural networks (DNNs) have achieved tremendous success and have been continuously applied in many application domains. However, during the practical deployment in industrial tasks, DNNs are found to be erroneous-prone due to various reasons such as overfitting and lacking of robustness to real-world corruptions during practical usage. To address these challenges, many recent attempts have been made to repair DNNs for version updates under practical operational contexts by updating weights (i.e., network parameters) through retraining, fine-tuning, or direct weight fixing at a neural level. Nevertheless, existing solutions often neglect the effects of neural network architecture and weight relationships across neurons and layers. In this work, as the first attempt, we initiate to repair DNNs by jointly optimizing the architecture and weights at a higher (i.e., block level). We first perform empirical studies to investigate the limitation of whole network-level and layer-level repairing, which motivates us to explore a novel repairing direction for DNN repair at the block level. To this end, we need to further consider techniques to address two key technical challenges, i.e., block localization , where we should localize the targeted block that we need to fix; and how to perform joint architecture and weight repairing . Specifically, we first propose adversarial-aware spectrum analysis for vulnerable block localization that considers the neurons’ status and weights’ gradients in blocks during the forward and backward processes, which enables more accurate candidate block localization for repairing even under a few examples. Then, we further propose the architecture-oriented search-based repairing that relaxes the targeted block to a continuous repairing search space at higher deep feature levels. By jointly optimizing the architecture and weights in that space, we can identify a much better block architecture. We implement our proposed repairing techniques as a tool, named ArchRepair , and conduct extensive experiments to validate the proposed method. The results show that our method can not only repair but also enhance accuracy and robustness, outperforming the state-of-the-art DNN repair techniques.

<i>DeepPatch</i>: A Patching-Based Method for Repairing Deep Neural Networks

Interpretability Based Neural Network Repair

DeepRepair: Style-Guided Repairing for Deep Neural Networks in the Real-World Operational Environment

PatchNAS: Repairing DNNs in Deployment with Patched Network Architecture Search.

DeepRepair: Style-Guided Repairing for DNNs in the Real-world Operational Environment

VeRe: Verification Guided Synthesis for Repairing Deep Neural Networks.

DeepPatch: Maintaining Deep Learning Model Programs to Retain Standard Accuracy with Substantial Robustness Improvement

Repairing Deep Neural Networks Based on Behavior Imitation

DeepCNN: A Dual Approach to Fault Localization and Repair in Convolutional Neural Networks

Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology

Cross-Patch Graph Convolutional Network for Image Denoising.

NNrepair: Constraint-based Repair of Neural Network Classifiers

ArchRepair : Block-Level Architecture-Oriented Repairing for Deep Neural Networks

DeepCorrect: Correcting DNN Models Against Image Distortions

Repairing Deep Neural Networks: Fix Patterns and Challenges

SpecRepair: Counter-Example Guided Safety Repair of Deep Neural Networks

Defensive Patches for Robust Recognition in the Physical World

Defending Adversarial Patches via Joint Region Localizing and Inpainting

DNNAM: Image inpainting algorithm via deep neural networks and attention mechanism

Benchmarking Adversarial Patch Against Aerial Detection

Training Patch Analysis and Mining Skills for Image Restoration Deep Neural Networks