Agent-based computational economics (ACE) combines elements from economics and computer science. In this article, the focus is on the relation between the evolutionary technique that is used and the economic problem that is modeled. In the field of ACE, economic simulations often derive parameter settings for the genetic algorithm directly from the values of the economic model parameters.This article compares two important approaches that are dominating in ACE and shows that the above practice may hinder the performance of the genetic algorithm and thereby hinder agent learning. More specifically, it is shown that economic model parameters and evolutionary algorithm parameters should be treated separately by comparing the two widely used approaches to social learning with respect to their convergence properties and robustness. This leads to new considerations for the methodological aspects of evolutionary algorithm design within the field of ACE.

Agent-based computational economics (ACE) combines elements from economics and computer science. In this paper, we focus on the relation between the evolutionary technique that is used and the economic problem that is modeled. In the field of ACE, economic simulations often derive parameter settings for the evolutionary algorithm directly from the values of the economic model parameters. In this paper, we compare two important approaches that are dominating ACE research and show that the above practice may hinder the performance of the evolutionary algorithm and thereby hinder agent learning. More specifically, we show that economic model parameters and evolutionary algorithm parameters should be treated separately by comparing the two widely used approaches to social learning with respect to their convergence properties and robustness. This leads to new considerations for the methodological aspects of evolutionary algorithm design within the field of ACE. Copyright Springer 2006evolutionary algorithms, simulation,

No abstract

No abstract

Comparisons of various methods for solving stochastic control economic models can be done with Monte Carlo methods. These methods have been applied to simple one-state, one-control quadratic-linear tracking models; however, large outliers may occur in a substantial number of the Monte Carlo runs when certain parameter sets are used in these models. Building on the work of Mizrach (1991) and Kendrick (1994, 1995), this paper tracks the source of these outliers to two sources: (1) the use of a zero for the penalty weights on the control variables and (2) the generation of near-zero initial estimate of the control parameter in the systems equations by the Monte Carlo routine. This result leads to an understanding of why both the unsophisticated Optimal Feedback (Certainty Equivalence) and the sophisticated Dual methods do poorly in some Monte Carlo comparisons relative to the moderately sophisticated Expected Optimal Feedback method.

stochastic control, economic stochastic control models, classification system, naming system, adaptive control, min–max control, robust control,

No abstract

Recently we have discovered an error in the implementation of the mutation operator in our earlier work on robust evolutionary algorithm design for socio-economic simulation (Alkemade et al. 2006(Alkemade et al. , 2007. 1 The original paper compared two commonly used approaches to socio-economic simulation. In the first approach parameter settings for the evolutionary algorithm are directly derived from the underlying economic model while in the second approach to social learning parameter settings are chosen so as to optimise evolutionary algorithm performance. Main conclusions of the original paper are that the first approach may hinder the performance of the evolutionary algorithm and thereby hinder agent learning, that is, that social learning evolutionary algorithms are able to overcome the so-called spite-effect and obtain high profit outcomes. These main conclusions are still confirmed when the error in the mutation operator is corrected. However, the convergence behaviour of some of the individual runs differs significantly from the (incorrect) results presented in the earlier papers. More specifically, in the corrected experiments we do not observe the same type of premature convergence in approach I. In this paper we present the corrected results. The average convergence behaviour for the two approaches is shown in Fig. 1, where we see convergence to the higher profit Cournot Nash outcome (at output 40) using approach II whereas approach I leads to the lower-profit competitive outcome (at output 50) for these set of EA parameters. The corrected results for the individual runs are shown in Fig.