Multi-Scale Dense Convolutional Networks for Efficient Prediction

This paper studies convolutional networks that require limited computational resources at test time. We develop a new network architecture that performs on par with state-of-the-art convolutional networks, whilst facilitating prediction in two settings: (1) an anytime-prediction setting in which the network's prediction for one example is progressively updated, facilitating the output of a prediction at any time; and (2) a batch computational budget setting in which a fixed amount of computation is available to classify a set of examples that can be spent unevenly across 'easier' and 'harder' examples. Our network architecture uses multi-scale convolutions and progressively growing feature representations, which allows for the training of multiple classifiers at intermediate layers of the network. Experiments on three image-classification datasets demonstrate the efficacy of our architecture, in particular, when measured in terms of classification accuracy as a function of the amount of compute available.