Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Accelerating Hybrid Agent-Based Models and Fuzzy Cognitive Maps: How to Combine Agents who Think Alike?

Philippe J. Giabbanelli, Jack T. Beerman

TL;DR

This paper presents an approximation that combines agents who ‘think alike’, thus reducing the population size and the compute time, and relies on representing agent behaviors as networks of rules and empirically evaluating different measures of distance between these networks.

Abstract

While Agent-Based Models can create detailed artificial societies based on individual differences and local context, they can be computationally intensive. Modelers may offset these costs through a parsimonious use of the model, for example by using smaller population sizes (which limits analyses in sub-populations), running fewer what-if scenarios, or accepting more uncertainty by performing fewer simulations. Alternatively, researchers may accelerate simulations via hardware solutions (e.g., GPU parallelism) or approximation approaches that operate a tradeoff between accuracy and compute time. In this paper, we present an approximation that combines agents who `think alike', thus reducing the population size and the compute time. Our innovation relies on representing agent behaviors as networks of rules (Fuzzy Cognitive Maps) and empirically evaluating different measures of distance between these networks. Then, we form groups of think-alike agents via community detection and simplify them to a representative agent. Case studies show that our simplifications remain accuracy.

Accelerating Hybrid Agent-Based Models and Fuzzy Cognitive Maps: How to Combine Agents who Think Alike?

TL;DR

This paper presents an approximation that combines agents who ‘think alike’, thus reducing the population size and the compute time, and relies on representing agent behaviors as networks of rules and empirically evaluating different measures of distance between these networks.

Abstract

While Agent-Based Models can create detailed artificial societies based on individual differences and local context, they can be computationally intensive. Modelers may offset these costs through a parsimonious use of the model, for example by using smaller population sizes (which limits analyses in sub-populations), running fewer what-if scenarios, or accepting more uncertainty by performing fewer simulations. Alternatively, researchers may accelerate simulations via hardware solutions (e.g., GPU parallelism) or approximation approaches that operate a tradeoff between accuracy and compute time. In this paper, we present an approximation that combines agents who `think alike', thus reducing the population size and the compute time. Our innovation relies on representing agent behaviors as networks of rules (Fuzzy Cognitive Maps) and empirically evaluating different measures of distance between these networks. Then, we form groups of think-alike agents via community detection and simplify them to a representative agent. Case studies show that our simplifications remain accuracy.
Paper Structure (14 sections, 2 figures, 6 tables, 1 algorithm)

This paper contains 14 sections, 2 figures, 6 tables, 1 algorithm.

Table of Contents

  1. INTRODUCTION
  2. BACKGROUD
  3. A Brief Introduction to Hybrid Agent-Based Models and Fuzzy Cognitive Maps
  4. Comparing Behaviors via Networks: New and Established Measures
  5. METHODS
  6. First Step: Weighing Social Ties by the Expected Similarity of Behavioral Rules
  7. Second Step: Identify Groups of Like-Minded Agents by Clustering the Weighted Social Ties
  8. Third Step: Aggregating Each Cluster into a Representative Super-Agent
  9. RESULTS
  10. Overview
  11. Case Studies
  12. Joint Effect of Community Detection Algorithm, Similarity Metric, and Network Structure
  13. Fidelity of Simulation Results Between the Simplified and the Original Model
  14. Discussion and Conclusion

Figures (2)

  • Figure 1: A hybrid ABM/FCM consists of agents who interact with each other (e.g., $A$ interacts with $B$ and each one has three other neighbors). Interactions are impacted by their 'mental model' in the form of an FCM, which is a network that performs simulations. As exemplified for agent $B$, simulating an FCM changes its node values (in the interval [0, 1]) over discrete iterations. To compare mental models, we can compare FCMs by transforming these networks into distributions (e.g., degree distribution) or vector embeddings, or on the basis of simple criteria such as the number of nodes. Each existing social tie is then weighted to reflect the similarity of mental models. Here, similarity in number of nodes is about 0.07.
  • Figure 2: In the obesity case study, mean simulation outcomes in the original model are comparable with the simplified model. However, simplified models have more uncertainty, as shown by their wider distributions.