Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Nash Equilibrium Learning In Large Populations With First-Order Payoff Modifications

Matthew S. Hankins, Jair Certório, Tzuyu Jeng, Nuno C. Martins

TL;DR

This paper develops a unified framework for Nash equilibrium learning in large populations under first-order payoff modifications. It couples two nonstandard system-theoretic notions—counterclockwise dissipativity (CCW) and delta-passivity—to prove convergence for a broad class of payoff mechanisms and learning dynamics, including discontinuous best-response and continuous revisions. The analysis introduces an evolutionary differential inclusion (EDIM) blending best-response with continuous learning rules and proves that, under certain passivity conditions and boundedness assumptions, population trajectories converge to the Nash equilibrium set of an associated stationary game. A numerical example based on Braess's paradox demonstrates practical toll design that steers equilibria toward more efficient outcomes. The results extend prior work by enabling joint CCW/delta-passivity treatment of first-order payoff modifications and guiding design of payoff mechanisms in large populations.

Abstract

We establish Nash equilibrium learning in large populations of noncooperative, strategic agents. Our analysis considers the broadest class to date of payoff mechanisms with first-order modifications, capable of modeling bounded rationality and anticipatory effects, averaging, or Padé delay approximations. We propose a framework that, for the first time, combines two nonstandard system-theoretic passivity notions. Our results hold for discontinuous best response dynamics alongside continuous learning rules, significantly extending prior work.

Nash Equilibrium Learning In Large Populations With First-Order Payoff Modifications

TL;DR

This paper develops a unified framework for Nash equilibrium learning in large populations under first-order payoff modifications. It couples two nonstandard system-theoretic notions—counterclockwise dissipativity (CCW) and delta-passivity—to prove convergence for a broad class of payoff mechanisms and learning dynamics, including discontinuous best-response and continuous revisions. The analysis introduces an evolutionary differential inclusion (EDIM) blending best-response with continuous learning rules and proves that, under certain passivity conditions and boundedness assumptions, population trajectories converge to the Nash equilibrium set of an associated stationary game. A numerical example based on Braess's paradox demonstrates practical toll design that steers equilibria toward more efficient outcomes. The results extend prior work by enabling joint CCW/delta-passivity treatment of first-order payoff modifications and guiding design of payoff mechanisms in large populations.

Abstract

We establish Nash equilibrium learning in large populations of noncooperative, strategic agents. Our analysis considers the broadest class to date of payoff mechanisms with first-order modifications, capable of modeling bounded rationality and anticipatory effects, averaging, or Padé delay approximations. We propose a framework that, for the first time, combines two nonstandard system-theoretic passivity notions. Our results hold for discontinuous best response dynamics alongside continuous learning rules, significantly extending prior work.

Paper Structure

This paper contains 20 sections, 5 theorems, 28 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Framework and Evolutionary Dynamics
  3. Learning Rules And Evolutionary Dynamics
  4. Best response rule and associated evolutionary dynamics
  5. Continuous rules and associated evolutionary dynamics
  6. Evolutionary differential inclusion
  7. Correlation function and properties
  8. Well-behaved rule classes and Examples
  9. Payoff Mechanism And Nash Equilibrium
  10. Solutions and a Key Assumption
  11. Potential and Contractive Games
  12. Linear Time Invariant Payoff Mechanism
  13. Nash Equilibria Set And Best Response Map
  14. System-theoretic Passivity Concepts
  15. Counterclockwise dissipativity (CCW)
  16. ...and 5 more sections

Key Result

Proposition 1

For all $x$ in $\mathbb{X}$ and $p$ in $\mathbb{R}^n$, it holds that

Figures (3)

  • Figure 1: (EDIM) in feedback with payoff mechanism $\mathfrak{P}=\mathfrak{S}\circ\mathfrak{F}$.
  • Figure 2: Traffic network and latency functions alongside the links.
  • Figure 3: Plot showing convergence for three distinct learning rule-initial condition pairs. Trajectories with the best response term are dotted.

Theorems & Definitions (23)

  • Proposition 1
  • Definition 1
  • Proposition 2
  • Example 1
  • Definition 2
  • Example 2
  • Definition 3
  • Example 3
  • Definition 4
  • Definition 5
  • ...and 13 more