On the Robustness of Convolutional Neural Networks to Internal Architecture and Weight Perturbations
This work addresses the problem of understanding network stability for researchers and practitioners in deep learning, though it is incremental as it extends robustness analysis from external to internal factors.
The paper investigates the robustness of convolutional neural networks to internal perturbations in weights and architecture, finding that higher layers are surprisingly robust (e.g., AlexNet shows less than 30% performance decrease with over 70% weight removal in top layers), while bottom layers are fragile (performance drops to near chance with similar perturbations).
Deep convolutional neural networks are generally regarded as robust function approximators. So far, this intuition is based on perturbations to external stimuli such as the images to be classified. Here we explore the robustness of convolutional neural networks to perturbations to the internal weights and architecture of the network itself. We show that convolutional networks are surprisingly robust to a number of internal perturbations in the higher convolutional layers but the bottom convolutional layers are much more fragile. For instance, Alexnet shows less than a 30% decrease in classification performance when randomly removing over 70% of weight connections in the top convolutional or dense layers but performance is almost at chance with the same perturbation in the first convolutional layer. Finally, we suggest further investigations which could continue to inform the robustness of convolutional networks to internal perturbations.