Distributed Optimization Environment for Bioelectromagnetism

Abstract:
This paper describes a distributed optimization system designed to use hybrid local area network resources to solve computationally expensive inverse problems of bioelectromagnetism. The system consists of a numerical core which extends the scientific library for evolutionary computations, and a network communication module built on Java RMI technology. It is portable, effective and relatively inexpensive. The authors present an example usage of the system for optimal design of a magnetic stimulator for vagus nerve. The numerical model of the nerve is discussed to choose the proper criteria for optimal design. The experiments have shown the linear scalability of the system and its good overall performance.