Design of an Incremental Machine Learning Component of a Ubiquitous Multimodal Multimedia Computing System

Machine learning (ML) allows a machine to acquire basic and decision-making knowledge so that it can be delegated with tasks that are otherwise left to human for implementation. Incremental learning is a form of continuous ML wherein the knowledge acquisition can go on indefinitely for as long as there is some new knowledge to learn. In effect, a machine becomes smarter after gaining added knowledge from training. Our paper details the design of an incremental ML component of a ubiquitous multimodal multimedia computing system. Our system chooses the media and modalities that are appropriate for the user situation; the situation itself is based upon user's context, user's profile and the user's environment and hardware profile. The user context itself is dependent on the user's location, the noise level in the workplace and the presence or absence of other people in the vicinity, and the user's computing device. If applicable, the user's special need (i.e. handicap) is also taken into consideration in the media and modality selection. Incremental ML allows our system to become fault-tolerant, capable of finding replacement to a missing or defective component