A Multi-Feature Pattern Recognition for P2P-based System Using In-network Associative Memory

Associative memory enables recall through invoking associations between past experiences in memory. In this paper, we demonstrate our fully distributed associative memory approach, the Distributed Associative Memory Tree (DASMET), to deal with multi-feature recognition in a peer-to-peer(P2P)-based system. The scheme constructs logical tree like structures within a peer-to-peer network enabling the nodes to search for patterns comprising multiple temporal or spatial features within a fixed number of steps using in-network processing. In doing so, the information held at individual peers is integrated into a common knowledge base which can be associatively searched by any peer within the network. We show that our scheme is fault tolerent and incurs low complexity overhead. By comparing our scheme to Backpropagation network and Radial Basis Function (RBF) network on two standard datasets, we prove its scalability and accuracy. Practically, this work has many advantages for the P2P domain. For example, various aspects such as file piracy issues, sensitive data leaking, and content pollution problem may be addressed.