Comparison study between conventional machine learning and distributed multi-task learning models

摘要

Applying machine learning in IoT devices is a challenge due to various reasons, such as the tremendous amount of data generated from IoT, the limitation of IoT devices' resources, and the non-IID nature of IoT data. On the other hand, transferring the generated IoT data to the cloud to train machine learning models consumes a lot of Bandwidth. Applying the distributed learning aspect in IoT large-scale deployments solves such issues, by employing edge computing devices as local cloud models in each location. This solution enhances the network overhead and helps in obtaining general models. However, this comes at the expense of the accuracy of the generated models. This paper provides a comparison study between applying a conventional machine learning model with a distributed multi-task learning model and discusses the factors that affect the distributed multi-task learning model.