Mimicking Ensemble Learning with Deep Branched Networks
Computer Vision and Pattern Recognition
2017-02-22 v1
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.
Cite
@article{arxiv.1702.06376,
title = {Mimicking Ensemble Learning with Deep Branched Networks},
author = {Byungju Kim and Youngsoo Kim and Yeakang Lee and Junmo Kim},
journal= {arXiv preprint arXiv:1702.06376},
year = {2017}
}