Hierarchical Training of Deep Neural Networks Using Early Exiting

Deep Neural Networks provide state-of-the-art accuracy for vision tasks but they require significant resources for training. Thus, they are trained on cloud servers far from the edge devices that acquire the data. This issue increases communication cost, runtime and privacy concerns. In this study, a novel hierarchical training method for deep neural networks is proposed that reduces the communication cost, training runtime, and privacy concerns by dividing the architecture between edge and cloud workers using early exits. The method proposes a brand-new use case for early exits to separate the backward pass of neural networks between the edge and the cloud during the training phase. We address the issues of most available hierarchical training methods that due to the sequential nature of the training phase, cannot train the levels of hierarchy at the same time or they do it with the cost of privacy. In contrast to these schemes, our method can use both edge and cloud workers simultaneously, does not share the raw input data with the cloud, and does not require communication during the backward pass. Several simulations and on-device experiments for different neural network architectures are done to demonstrate the effectiveness of this method. It is shown that the method reduces 29% and 61% runtime in CIFAR-10 classification experiment for VGG-16 and ResNet-18 when the communication with the cloud is done over the 3G protocol. This gain in the runtime is achieved whilst the accuracy drop is negligible. This method can be inspirational to provide online learning of high-accuracy deep neural networks on low-resource devices such as mobile phones or robots as a part of an edge-cloud system, making them more flexible in facing new tasks and classes of data in the future.