Abstract
Rich phenomena from complex systems have long intrigued researchers, and yet modeling system micro-dynamics and inferring the forms of interaction remain challenging for conventional data-driven approaches, being generally established by scientists with human ingenuity. In this study, we propose AgentNet, a model-free data-driven framework consisting of deep neural networks to reveal and analyze the hidden interactions in complex systems from observed data alone. AgentNet utilizes a graph attention network with novel variable-wise attention to model the interaction between individual agents, and employs various encoders and decoders that can be selectively applied to any desired system. Our model successfully captured a wide variety of simulated complex systems, namely cellular automata (discrete), the Vicsek model (continuous), and active Ornstein–Uhlenbeck particles (non-Markovian) in which, notably, AgentNet’s visualized attention values coincided with the true variable-wise interaction strengths and exhibited collective behavior that was absent in the training data. A demonstration with empirical data from a flock of birds showed that AgentNet could identify hidden interaction ranges exhibited by real birds, which cannot be detected by conventional velocity correlation analysis. We expect our framework to open a novel path to investigating complex systems and to provide insight into general process-driven modeling.
Highlights
Complex systems are collections of interactive agents that exhibit non-trivial collective behavior
Several attempts have been separately made to employ graph neural network (GNN) in the prediction and analysis of specific complex systems and physical m odels[19,20,21,22,23,24,25,26,27,28,29], but these approaches are mostly limited to the verification of a single system or a small number of agents, and more significantly, it remains difficult to interpret the characteristics of the interaction due to the neural network’s notorious black-box nature
This study proposed AgentNet, a generalized framework for the data-driven modeling of a complex system
Summary
Complex systems are collections of interactive agents that exhibit non-trivial collective behavior. In a multi-dimensional system, the interaction strength cannot be a simple scalar value since each state variable possesses its own interaction range and strength and may yield an assymteric, inhomogeneous interaction range Inspired by these recent attempts, we introduce AgentNet, a generalized neural network framework based on GATs with a novel attention scheme to model various complex systems in an physically interpretable manner. Our model jointly learns the interaction strength that affects each variable’s transition and overall transition function from observed data in an end-to-end manner without any human intervention or manual operation This is a critical difference from the conventional approach with GATs, which only assigns a single attention value per agent while our model assigns completely independent attention values for every state variable and employing separate decoders for each of them. A trained AgentNet can generate an individual level of state predictions from desired initial conditions, making AgentNet an outstanding simulator of target systems including even those that exhibit collective behavior that was absent in the training data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
