Geometric robustness of deep networks: analysis and improvement
This addresses security concerns in sensitive applications where geometric robustness is critical, though it builds incrementally on existing adversarial training methods.
The authors tackled the problem of deep neural networks' vulnerability to geometric transformations by proposing ManiFool, an algorithm to measure invariance, and an adversarial training scheme to improve it. They demonstrated ManiFool's scalability on complex networks and high-dimensional datasets, showing effectiveness in enhancing robustness.
Deep convolutional neural networks have been shown to be vulnerable to arbitrary geometric transformations. However, there is no systematic method to measure the invariance properties of deep networks to such transformations. We propose ManiFool as a simple yet scalable algorithm to measure the invariance of deep networks. In particular, our algorithm measures the robustness of deep networks to geometric transformations in a worst-case regime as they can be problematic for sensitive applications. Our extensive experimental results show that ManiFool can be used to measure the invariance of fairly complex networks on high dimensional datasets and these values can be used for analyzing the reasons for it. Furthermore, we build on Manifool to propose a new adversarial training scheme and we show its effectiveness on improving the invariance properties of deep neural networks.