Coupled Ensembles of Neural Networks

We investigate in this paper architectures of deep convolutional networks. Building on existing state of the art models, we propose a reconfiguration of the model parameters into several parallel branches at the global network level, each branch being a standalone CNN. We show that this arrangement is an efficient way to significantly reduce the number of parameters while at the same time improving the performance. The use of branches brings an additional form of regularization. In addition to splitting the parameters into parallel branches, we propose a tighter coupling of these branches by averaging their log-probabilities. The tighter coupling favours the learning of better representations, even at the level of the individual branches, as compared to when each branch is trained independently. We refer to this branched architecture as “coupled ensembles”. The approach is generic and can be applied to almost any neural network architecture. With coupled ensembles of DenseNet-BC networks and a 25M-parameter budget, we obtain error rates of 2.92%, 15.68% and 1.50% on CIFAR-10, CIFAR-100 and SVHN respectively. For the same parameter budget, DenseNet-Bchas an error rate of 3.46%, 17.18%, and 1.8% respectively. With ensembles of coupled ensembles of DenseNet-BC networks with a 50M-parameter budget, we obtain error rates of 2.72%, 15.13% and 1.42 % respectively on these tasks.