Ensemble learning in CNN augmented with fully connected subnetworks
This work addresses performance enhancement in object recognition for researchers and practitioners, but it is incremental as it builds on existing CNN architectures with a novel ensemble method.
The paper tackles improving CNN performance by proposing EnsNet, which combines a base CNN with multiple fully connected subnetworks (FCSNs) trained independently on disjoint feature-map subsets and uses majority voting for predictions, achieving a state-of-the-art error rate of 0.16% on MNIST.
Convolutional Neural Networks (CNNs) have shown remarkable performance in general object recognition tasks. In this paper, we propose a new model called EnsNet which is composed of one base CNN and multiple Fully Connected SubNetworks (FCSNs). In this model, the set of feature-maps generated by the last convolutional layer in the base CNN is divided along channels into disjoint subsets, and these subsets are assigned to the FCSNs. Each of the FCSNs is trained independent of others so that it can predict the class label from the subset of the feature-maps assigned to it. The output of the overall model is determined by majority vote of the base CNN and the FCSNs. Experimental results using the MNIST, Fashion-MNIST and CIFAR-10 datasets show that the proposed approach further improves the performance of CNNs. In particular, an EnsNet achieves a state-of-the-art error rate of 0.16% on MNIST.