Scalable Smartphone Cluster for Deep Learning
This work addresses the computational bottleneck for deep learning training on smartphone clusters, though it is incremental as it removes portability to improve efficiency.
The authors tackled the problem of insufficient computational power for training deep neural networks on small-scale smartphone clusters by proposing a scalable cluster using 138 wired Galaxy S10+ devices, achieving 90% of a P100's speed for ResNet-50 and a 43x speed-up over a V100 for MobileNet-v1.
Various deep learning applications on smartphones have been rapidly rising, but training deep neural networks (DNNs) has too large computational burden to be executed on a single smartphone. A portable cluster, which connects smartphones with a wireless network and supports parallel computation using them, can be a potential approach to resolve the issue. However, by our findings, the limitations of wireless communication restrict the cluster size to up to 30 smartphones. Such small-scale clusters have insufficient computational power to train DNNs from scratch. In this paper, we propose a scalable smartphone cluster enabling deep learning training by removing the portability to increase its computational efficiency. The cluster connects 138 Galaxy S10+ devices with a wired network using Ethernet. We implemented large-batch synchronous training of DNNs based on Caffe, a deep learning library. The smartphone cluster yielded 90% of the speed of a P100 when training ResNet-50, and approximately 43x speed-up of a V100 when training MobileNet-v1.