Speeding Coordination by Combining Analytical and Inductive Learning

In a highly dynamic information society, the practical applicability of one filtering framework is usually directly proportional to its flexibility, where this assumes not only an easy integration of novel strategies but also the ability to adapt to new resource conditions. A major drawback of many existing systems, trying to make different synergies between filtering strategies, is usually concerned with an inability to easily integrate new strategies and with not taking care of resource availability, being critical for the realisation of the successful commercial deployments. The cornerstone of the presented filtering framework is in the encapsulation of the searching algorithms inside separate filtering agents whose abilities to be utilised in different runtime situations are efficiently learnt by combining both analytical and inductive learning. The evaluation results demonstrate that analytical learning successfully exploits domain knowledge about filtering strategies while helping inductive learning do faster adaptation.