Mimicking Ensemble Learning with Deep Branched Networks

TL;DR

This paper introduces a branched residual network that mimics ensemble learning within a single model, sharing low-level features to improve image classification performance on ImageNet.

Contribution

It presents a novel deep branched residual network architecture that replicates ensemble learning effects within a single network, enhancing classification accuracy.

Findings

Achieved improved ImageNet classification accuracy

Demonstrated effective sharing of low-level features

Mimicked ensemble benefits within a single network

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.