Many machine-learning techniques use feedback information. However, current context fusion systems do not support this because they constrain processing to be structured as acyclic processing. This paper proposes a generalization which enables the use of cyclic processing in context fusion systems. A solution is proposed to the inherent problem of how to avoid uncontrollable looping during cyclic processing. The solution is based on finding cycles using graph-coloring and breaking cycles using time constraints.
Adjunct Proceedings of the Fifth International Conference on Pervasive Computing (pervasive 2007), 2007, p. 41-44
Main Research Area:
The Fifth International Conference on Pervasive Computing, 2007