A multiple-transfer framework for learning context models for dynamic smart-home environments

A real living space is dynamic in nature, which leads to various changes over time and would render a smart-home system incapable of providing reliable services. In this regard, a smart home often needs to keep context models adaptable, which may cause tremendous efforts. To reduce the efforts, a multiple-transfer framework is proposed to transfer knowledge from a source domain to a target one by reusing as much information from the source domain. This way, the efforts of model training in the target domain can be effectively reduced. The proposed framework provides flexibility of replacing its internal components to help a smart home respond to inevitable changes particularly for transferring knowledge to a new domain. Such design will improve the overall adaptability and practicality and the preliminary results also show the potentials of the framework.