In this thesis, we propose the architecture PIAF (Personnal Intelligent Agent Framework) whose objective is to provide users with an environment for nonintrusive, autonomous and general-purpose exchange of information. The problems of diffusion of information between users and optimization of the network's topology are approached with an algorithm using artificial ants. The use of artificial pheromones deposited on connexions between peers at the time of the transfers authorizes the constitution of a global memory of the exchanges and the detection of shared centers of interests. Comparatively with existing solutions, the advantage of our algorithm is to free the user from the definition of profiles. This last needs neither to subscribe with diffusion channel nor to define its centers of interests to be able to exchange information.
展开▼
