# On the Robustness of Convolutional Neural Networks to Internal Architecture and Weight Perturbations

arXiv: https://arxiv.org/abs/1703.08245 · HTML: https://mlsanity.com/paper/1703.08245/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 23 Mar 2017.

## Built on by (most cited first, 14 of 14)

- [Relating Adversarially Robust Generalization to Flat Minima](https://mlsanity.com/paper/2104.04448/) (2021)
- [Random and Adversarial Bit Error Robustness: Energy-Efficient and Secure DNN Accelerators](https://mlsanity.com/paper/2104.08323/) (2021)
- [WeiPer: OOD Detection using Weight Perturbations of Class Projections](https://mlsanity.com/paper/2405.17164/) (2024)
- [Adversarial Robustness in Deep Learning: Attacks on Fragile Neurons](https://mlsanity.com/paper/2201.12347/) (2022)
- [Investigating Weight-Perturbed Deep Neural Networks With Application in Iris Presentation Attack Detection](https://mlsanity.com/paper/2311.12764/) (2023)
- [Non-Singular Adversarial Robustness of Neural Networks](https://mlsanity.com/paper/2102.11935/) (2021)
- [Topological derivative approach for deep neural network architecture adaptation](https://mlsanity.com/paper/2502.06885/) (2025)
- [Combining Gradients and Probabilities for Heterogeneous Approximation of Neural Networks](https://mlsanity.com/paper/2208.07265/) (2022)
- [Exploring and Exploiting the Asymmetric Valley of Deep Neural Networks](https://mlsanity.com/paper/2405.12489/) (2024)
- [Mental Models of Adversarial Machine Learning](https://mlsanity.com/paper/2105.03726/) (2021)
- [Neuronal diversity can improve machine learning for physics and beyond](https://mlsanity.com/paper/2204.04348/) (2022)
- [Beyond topography: Topographic regularization improves robustness and reshapes representations in convolutional neural networks](https://mlsanity.com/paper/2508.00043/) (2025)
- [On the Stability of Growth in Structural Plasticity: Forward-Active yet Backward-Starved](https://mlsanity.com/paper/2605.15435/) (2026)
- [Understanding Fault Tolerance of Adversarially Robust Pruned Models](https://mlsanity.com/paper/2608.04173/) (2026)

## Abstract (arXiv)

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.