Coupled Ensembles of Neural Networks

Anuvabh Dutt (1) , Georges Quénot (1) , Denis Pellerin (2)

1 LIG - Laboratoire d'Informatique de Grenoble
2 GIPSA-AGPIG - GIPSA - Architecture, Géométrie, Perception, Images, Gestes

We present coupled ensembles of neural networks, which is a reconfiguration of existing neural network models into parallel branches. We empirically show that this modification leads to results on CIFAR and SVHN that are competitive to state of the art, with a greatly reduced parameter count. Additionally, for a fixed parameter, or a training time budget coupled ensembles are significantly better than single branch models. Preliminary results on ImageNet are also promising. Code for the experiments can be found at: https://github.com/vabh/ coupled_ensembles

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Recherche d'information [cs.IR]; Apprentissage [cs.LG]

Anuvabh Dutt, Georges Quénot, Denis Pellerin. Coupled Ensembles of Neural Networks. ICLR 2018 - 6th International Conference on Learning Representations - Workshop Track, Apr 2018, Vancouver, Canada. ⟨hal-01887644⟩