IC Networks: Remodeling the Basic Unit for Convolutional Neural Networks
This work provides an incremental improvement in CNN architecture design, offering a more efficient basic unit for computer vision practitioners working with large-scale image classification.
This paper introduces the Inter-layer Collision (IC) structure, a new basic unit for Convolutional Neural Networks (CNNs) inspired by the elastic collision model. When integrated into ResNet-50, it reduced the top-1 error on ImageNet from 22.38% to 21.75%, matching ResNet-100's performance with approximately half the FLOPs.
Convolutional neural network (CNN) is a class of artificial neural networks widely used in computer vision tasks. Most CNNs achieve excellent performance by stacking certain types of basic units. In addition to increasing the depth and width of the network, designing more effective basic units has become an important research topic. Inspired by the elastic collision model in physics, we present a general structure which can be integrated into the existing CNNs to improve their performance. We term it the "Inter-layer Collision" (IC) structure. Compared to the traditional convolution structure, the IC structure introduces nonlinearity and feature recalibration in the linear convolution operation, which can capture more fine-grained features. In addition, a new training method, namely weak logit distillation (WLD), is proposed to speed up the training of IC networks by extracting knowledge from pre-trained basic models. In the ImageNet experiment, we integrate the IC structure into ResNet-50 and reduce the top-1 error from 22.38% to 21.75%, which also catches up the top-1 error of ResNet-100 (21.75%) with nearly half of FLOPs.