Angielski
This book frames a peer-to-peer information retrieval§problem as a multi-agent framework and attacks it§from an organizational perspective by exploring§various adaptive, self-organizing topological§organizations, designing appropriate§coordination strategies, and exploiting learning§techniques to create more accurate routing policy for§large-scale agent organizations. In addition, a§reinforcement-learning based approach is developed in§this thesis to take advantage of the run-time§characteristics of P2P IR systems, including§environmental parameters, bandwidth usage, and§historical information about past search sessions. In§the learning process, agents refine their content§routing policies by constructing relatively accurate§routing tables based on a Q-learning algorithm.§Experimental results show that this learning§algorithm considerably improves the performance of§distributed search sessions in P2P IR systems.§§The book is addressed to researchers and§practitioners in information retrieval and search§engine, content-based routing areas.
