Committees of deep feedforward networks trained with few data
This addresses the problem of data scarcity in image classification for researchers, though it appears incremental.
The authors tackled image classification with limited training data by combining multiple deep convolutional neural networks in a committee, achieving results better than state-of-the-art on the STL-10 dataset.
Deep convolutional neural networks are known to give good results on image classification tasks. In this paper we present a method to improve the classification result by combining multiple such networks in a committee. We adopt the STL-10 dataset which has very few training examples and show that our method can achieve results that are better than the state of the art. The networks are trained layer-wise and no backpropagation is used. We also explore the effects of dataset augmentation by mirroring, rotation, and scaling.