Attribute-driven Design of an Incremental Learning Component of a Ubiquitous Multimodal Multimedia Computing System

M. D. Hina, C. Tadj, A. R. Cherif

Our ubiquitous multimodal multimedia computing system selects appropriate media and modalities based on user’s context, user’s profile and environment profile. The user’ context and profile constitute a pre-condition scenario. The appropriate media and modalities constitute the post-condition scenario. The machine learning system detects the user scenario based on its knowledge database. In general, the post-condition scenario is to be adapted for implementation. Whenever there is a media or modality listed in the post-condition scenario that is missing or defective, the system would find alternative device/modality from its knowledge database. In the absence of list of alternatives or in case of cascaded failures of different media or modalities, the system would consult the user for the appropriate replacement for the failed components. This is incremental learning, the progressive acquisition of knowledge that allows the system to remain fault-tolerant. This paper demonstrates the design of the machine learning component using attribute-driven design.