Mimicking Ensemble Learning with Deep Branched Networks

2017/02/21 by Byungju Kim, Youngsoo Kim, Kim, Byungju +5
Computer Science · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #Domain Adaptation and Few-Shot Learning #cs.CV

paper · pdf · doi:10.48550/arxiv.1702.06376

arxiv created 2017/02/21 · arxiv updated 2017/02/22

Abstract

This paper proposes a branched residual network for image classification. It is known that high-level features of deep neural network are more representative than lower-level features. By sharing the low-level features, the network can allocate more memory to high-level features. The upper layers of our proposed network are branched, so that it mimics the ensemble learning. By mimicking ensemble learning with single network, we have achieved better performance on ImageNet classification task.

Related

• Deep Network Ensemble Learning applied to Image Classification using CNN Trees
• Multi-Residual Networks: Improving the Speed and Accuracy of Residual Networks
• Deep Pyramidal Residual Networks
• ShaResNet: reducing residual network parameter number by sharing weights
• Aggregated Residual Transformations for Deep Neural Networks