IMEXnet: A Forward Stable Deep Neural Network
This work addresses robustness and field of view issues in deep learning for computer vision, offering a novel architecture that could enhance performance in tasks like semantic segmentation, though it appears incremental as it builds on existing network designs with a new method.
The authors tackled the challenges of robustness to input perturbations and limited field of view in deep convolutional neural networks by introducing IMEXnet, which adapts semi-implicit methods from partial differential equations, resulting in improved stability and generalization compared to explicit networks like residual networks.
Deep convolutional neural networks have revolutionized many machine learning and computer vision tasks, however, some remaining key challenges limit their wider use. These challenges include improving the network's robustness to perturbations of the input image and the limited ``field of view'' of convolution operators. We introduce the IMEXnet that addresses these challenges by adapting semi-implicit methods for partial differential equations. Compared to similar explicit networks, such as residual networks, our network is more stable, which has recently shown to reduce the sensitivity to small changes in the input features and improve generalization. The addition of an implicit step connects all pixels in each channel of the image and therefore addresses the field of view problem while still being comparable to standard convolutions in terms of the number of parameters and computational complexity. We also present a new dataset for semantic segmentation and demonstrate the effectiveness of our architecture using the NYU Depth dataset.