Convergence of learning algorithms without a projection facility

DOI10.1016/S0304-4068(97)00023-2MaRDI QIDQ1300379zbMATH OpenOpenAlexWikidataFDO

Authors George W. Evans, Seppo Honkapohja

Publication date 20 December 2000

Published in Journal of Mathematical Economics (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1016/s0304-4068(97)00023-2, https://www.cesifo.org/DocDL/ces_wp109.pdf

zbMATH Keywords

noisy sunspot equilibriarational expectationsstochastic approximation

Mathematics Subject Classification ID

Statistical methods; economic indices and measures (91B82) Stochastic learning and adaptive control (93E35)

Review text

This paper addresses convergence of learning algorithms without the use of a projection facility. In earlier research direct proofs of convergence in simple models was employed, but new research has showed that those same results an be obtained using the stochastic approximation technique of \textit{L. Ljung} [IEEE Trans. Autom. Control 22, 551-575 (1977; Zbl 0362.93031)]. This method is used to analyze local stability in stochastic dynamic economic models in the presence of multiple equilibria. The analysis is conducted under the assumption of Markovian state variable dynamics. Finally, we present two economic applications of the convergence result.

Recommendations

• Economic Dynamics with Learning: New Stability Results
• Two problems in applying Ljung's projection algorithms to the analysis of decentralized learning
• The convergence of least squares learning in stochastic temporary equilibrium models
• Local Convergence of Recursive Learning to Steady States and Cycles in Stochastic Nonlinear Models

Cites work

• A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle
• Adaptive Learning with Nonlinear Dynamics Driven by Dependent Processes
• Analysis of recursive stochastic algorithms
• Convergence of least squares learning mechanisms in self-referential linear stochastic models
• Economic Dynamics with Learning: New Stability Results
• scientific article; zbMATH DE number 3126031 (Why is no real title available?)
• scientific article; zbMATH DE number 19473 (Why is no real title available?)
• scientific article; zbMATH DE number 48727 (Why is no real title available?)
• scientific article; zbMATH DE number 53571 (Why is no real title available?)
• scientific article; zbMATH DE number 1154011 (Why is no real title available?)
• scientific article; zbMATH DE number 3276736 (Why is no real title available?)
• Indeterminacy of stationary equilibrium in stochastic overlapping generations models
• Invariant Distributions and the Limiting Behavior of Markovian Economic Models
• Learning to Believe in Sunspots
• Local Convergence of Recursive Learning to Steady States and Cycles in Stochastic Nonlinear Models
• On learning and rational expectations in an overlapping generations model
• On the local stability of sunspot equilibria under adaptive learning rules
• Self-fulfilling prophecies
• Strategic Complementarity in Business Formation: Aggregate Fluctuations and Sunspot Equilibria
• Sunspot-like effects of random endowments
• Two problems in applying Ljung's projection algorithms to the analysis of decentralized learning

Cited in

(4)

• Adaptive learning in practice
• Two problems in applying Ljung's projection algorithms to the analysis of decentralized learning
• Learning convergence of CMAC algorithm
• Coherence without rationality at the zero lower bound

This page was built for publication: Convergence of learning algorithms without a projection facility

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1300379)