Feature Fusion for Online Mutual Knowledge Distillation

TL;DR

This paper introduces Feature Fusion Learning (FFL), a framework that combines feature maps from parallel networks and mutually distills knowledge between sub-networks and a fused classifier, enhancing overall performance.

Contribution

The novel FFL framework enables online mutual knowledge distillation through feature fusion, improving accuracy of both sub-networks and the fused classifier across multiple datasets.

Findings

FFL outperforms existing methods on CIFAR-10, CIFAR-100, and ImageNet.

Mutual teaching improves individual sub-network performance.

Flexible network types can be used for sub-networks.

Abstract

We propose a learning framework named Feature Fusion Learning (FFL) that efficiently trains a powerful classifier through a fusion module which combines the feature maps generated from parallel neural networks. Specifically, we train a number of parallel neural networks as sub-networks, then we combine the feature maps from each sub-network using a fusion module to create a more meaningful feature map. The fused feature map is passed into the fused classifier for overall classification. Unlike existing feature fusion methods, in our framework, an ensemble of sub-network classifiers transfers its knowledge to the fused classifier and then the fused classifier delivers its knowledge back to each sub-network, mutually teaching one another in an online-knowledge distillation manner. This mutually teaching system not only improves the performance of the fused classifier but also obtains performance gain in each sub-network. Moreover, our model is more beneficial because different types of network can be used for each sub-network. We have performed a variety of experiments on multiple datasets such as CIFAR-10, CIFAR-100 and ImageNet and proved that our method is more effective than other alternative methods in terms of performance of both sub-networks and the fused classifier.