MASFENON: implementing a multi-agent simulation framework for interconnected networks with distributed programming

Abstract

The complexity of networked systems, particularly interconnected networks, necessitates advanced simulation frameworks to accurately emulate real-world dynamics, especially in the context of big data and high-performance computing. Most software used for simulation and temporal inference usually falls short in large data and optimization, since it is generally used in particular contexts, like simulating the dynamics of a specific group of entities, such as cellular and community interactions. We present ”Multi-Agent Adaptive Simulation Framework for Evolution in Networks of Networks” (MASFENON). MASFENON employs a temporal multi-layered approach to simulate and analyze dynamic processes in interconnected networks. The framework leverages parallel programming techniques for matrix and linear algebra operations and distributed and reactive programming for agent and environment communication, all implemented in C++ using the Message Passing Interface (MPI) standard. MASFENON has been validated against several common network models and could simulate the behavior of real systems in the context of epidemic simulations (See [4]). The framework demonstrates sublinear speedup and scalability with network size. The implementation is open source and available in a regularly updated GitHub repository1. MASFENON's integration of MPI and distributed programming techniques provides a powerful and versatile tool for modeling complex network interactions and dynamics. Its capabilities extend beyond traditional models, offering new insights and applications in network science.