PhD Thesis Defense Announcement-Cellular Learning Automata and Their Applications to Peer-to-Peer Networks-Ali Mohammad Saghiri

ABSTRACT- Cellular learning automata (CLAs) have received significant attentions by researchers in recent years. CLAs are learning models, which bring together the computational power of cellular automata (CAs), and the learning capability of learning automata (LAs). TheCLAs can be classified into two main classes: Static CLAs (SCLAs) and Dynamic CLAs (DCLAs). In a SCLA, the graph of the cellular structure remains fixed during the evolution of the CLA. Due to the dynamic nature of real world problems their structural properties are time varying and for this reason using fixed graphs for modeling them are too restrictive. In a DCLA, one of its aspects such as structure, local rule, or neighborhood may change over time. The DCLAs can be used to solve problems that can be modeled as dynamic graphs. In this dissertation, we first propose several models of DCLAs and then present an approach for designing cognitive peer-to-peer networks based on the proposed models. At first, in this dissertation, two models of DCLAs, namely Closed Asynchronous DCLA, and Closed Asynchronous DCLA with varying number of learning automata in each cell are proposed. Next, several cognitive engines based on the proposed models are presented for solving topology mismatch, and super-peer selection problems in peer-to-peer networks. To evaluate the proposed cognitive peer-to-peer networks several experiments have been conducted. Experimentations have shown that the proposed cognitive peer-to-peer networks for solving landmark clustering problem perform better than the existing algorithms with respect to communication delay. Also, experimentations have shown the superiority of the proposed cognitive peer-to-peer networks for solving super-peer selection problem over the existing algorithms with respect to capacity utilization and number of super-peers.

Attachment

Attachment

Publisher: Computer engineering| | Date: 2017/10/02