Convolution with even-sized kernels and symmetric padding
This work addresses efficiency and training challenges in compact convolutional neural networks, particularly for online and continual learning, though it is incremental as it builds on existing kernel methods.
The paper tackled the shift problem in even-sized kernel convolutions by proposing symmetric padding, which improved generalization and outperformed 3x3 kernels in image classification and generation tasks, with C2sp achieving comparable accuracy to compact models using less memory and time.
Compact convolutional neural networks gain efficiency mainly through depthwise convolutions, expanded channels and complex topologies, which contrarily aggravate the training process. Besides, 3x3 kernels dominate the spatial representation in these models, whereas even-sized kernels (2x2, 4x4) are rarely adopted. In this work, we quantify the shift problem occurs in even-sized kernel convolutions by an information erosion hypothesis, and eliminate it by proposing symmetric padding on four sides of the feature maps (C2sp, C4sp). Symmetric padding releases the generalization capabilities of even-sized kernels at little computational cost, making them outperform 3x3 kernels in image classification and generation tasks. Moreover, C2sp obtains comparable accuracy to emerging compact models with much less memory and time consumption during training. Symmetric padding coupled with even-sized convolutions can be neatly implemented into existing frameworks, providing effective elements for architecture designs, especially on online and continual learning occasions where training efforts are emphasized.