Sharing Residual Units Through Collective Tensor Factorization in Deep Neural Networks

Abstract

Residual units are wildly used for alleviating optimization difficulties when building deep neural networks. However, the performance gain does not well compensate the model size increase, indicating low parameter efficiency in these residual units. In this work, we first revisit the residual function in several variations of residual units and demonstrate that these residual functions can actually be explained with a unified framework based on generalized block term decomposition. Then, based on the new explanation, we propose a new architecture, Collective Residual Unit (CRU), which enhances the parameter efficiency of deep neural networks through collective tensor factorization. CRU enables knowledge sharing across different residual units using shared factors. Experimental results show that our proposed CRU Network demonstrates outstanding parameter efficiency, achieving comparable classification performance to ResNet-200 with the model size of ResNet-50. By building a deeper network using CRU, we can achieve state-of-the-art single model classification accuracy on ImageNet-1k and Places365-Standard benchmark datasets. (Code and trained models are available on GitHub)

References in corpus (17)

• Adam: A Method for Stochastic Optimization
• Deep Residual Learning for Image Recognition
• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
• Wide Residual Networks
• Tensorizing Neural Networks
• Residual Networks Behave Like Ensembles of Relatively Shallow Networks
• Speeding up Convolutional Neural Networks with Low Rank Expansions
• Aggregated Residual Transformations for Deep Neural Networks
• Speeding-up Convolutional Neural Networks Using Fine-tuned CP-Decomposition
• Pyramid Scene Parsing Network
• Wider or Deeper: Revisiting the ResNet Model for Visual Recognition
• Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
• Places: An Image Database for Deep Scene Understanding
• Ultimate tensorization: compressing convolutional and FC layers alike
• PolyNet: A Pursuit of Structural Diversity in Very Deep Networks
• Convolutional Rectifier Networks as Generalized Tensor Decompositions

Cited by in corpus (7)

• Deep Polynomial Neural Networks
• Image Super-Resolution via Dual-State Recurrent Networks
• Efficient Visual Recognition with Deep Neural Networks: A Survey on Recent Advances and New Directions
• Learning Compact Recurrent Neural Networks with Block-Term Tensor Decomposition
• Iteratively Training Look-Up Tables for Network Quantization
• BT-Nets: Simplifying Deep Neural Networks via Block Term Decomposition
• Block-term Tensor Neural Networks