Powering Smart Home Intelligence Using Existing Entertainment Systems

摘要

Smart Homes and Smart Environments are classically centrally organized and utilize methods from machine learning and artificial intelligence. In recent years, continuous progress has been made and the technology has reached a level where the deployment becomes feasible on a larger scale and in everyday settings, raising the question where the central system should be deployed. In this paper, we propose the use of existing entertainment systems as the Smart Home controller. In a case study we examine the performance of the multi-core Cell processor of the Sony PlayStation 3 for training artificial feed forward neural networks using specifically adapted parallel training strategies. The evaluation of these strategies shows a gain in speed of up to 6.6 over a sequential implementation on a single processing element of the Cell. Based on these findings and related work we suggest that home entertainment systems should be considered as possible powerful, deployment platforms for future Smart Home systems.