Category:Learning to readAffordable Access
An agent’s tool for inference in model-based design for parameter estimation and model selection
By
Mazhar A. Chaudhry
Published By
Elsevier Science
Article first published online:
30 July 2011
DOI:
10.1016/j.ijpara.2011.06.001
Abstract
Here we present a novel agent-based approach for model-based design (MBD) and inference of parameters of stochastic models (SMs) by a cooperative agent. We show that this agent-based MBD framework can be used to infer a range of model parameters like the ability of the environment to keep the agents well aligned with their desired trajectories, the cost and risk associated with acquiring any information, and the ability of the environment to recover from any perturbations. We show the agent’s tool for model selection by comparing the agent-based approach with a traditional black-box approach. Our results show that the agent-based MBD is better in finding the optimal number of agents needed to reach their goal in an environment. Furthermore, we also show that this agent-based MBD framework outperforms the traditional black-box model selection approach by a small margin in high-dimensional problems. This performance gain comes from the ability of the agent to infer relevant information about the dynamics of the SMs and to optimize its behaviour accordingly in a self-motivated manner.
Elsevier is a leading global information solutions company that helps professionals stay ahead of the curve. By providing trusted solutions and services, we help our customers advance their careers and improve their business performance. Elsevier provides web-based, application and multimedia publishing solutions; professional book publishing and journal publishing services; ad knowledge management and marketing solutions; a comprehensive materials service; and research solutions that include databases, subscriptions, and training. We also provide science solutions that include the world's most comprehensive digital media solutions, and a hosting, analysis, and informatics service that keeps our customers ahead of the curve. Elsevier is part of RELX Group plc (NYSE: RXX), a world-leading provider of information and analytics for professional and business customers across industries.
Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system. be359ba680
Related links:
Comments