IC Networks: Remodeling the Basic Unit for Convolutional Neural Networks

TL;DR

This paper introduces the Inter-layer Collision (IC) structure for CNNs, enhancing feature extraction and performance, and proposes weak logit distillation to accelerate training, demonstrated by improved results on ImageNet.

Contribution

The paper presents the IC structure as a novel CNN unit inspired by physics, and a new training method WLD, both improving CNN performance and training efficiency.

Findings

IC structure improves CNN accuracy on ImageNet

WLD accelerates training of IC networks

IC reduces FLOPs while maintaining accuracy

Abstract

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.