Building effective deep neural network architectures one feature at a time

Martin Mundt, Tobias Weis, Kishore Konda +1

Successful training of convolutional neural networks is often associated with sufficiently deep architectures composed of high amounts of features. These networks typically rely on…