Deep Reinforcement Learning: Emerging Trends in Macroeconomics and Future Prospects
IMF Working Papers, December 16, 2022
Source details
- Canonical URL
- Deep Reinforcement Learning: Emerging Trends in Macroeconomics and Future Prospects
Other formats
Bibliographic details
- Authors: Tohid Atashbar, Rui Aruhan Shi
- Published: December 16, 2022
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9798400224713.001
Overview and purpose
- Authors: Tohid Atashbar, Rui Aruhan Shi
- Publication date: December 16, 2022
- Series: Working Paper No. 2022/259
- Issue: 259
- Volume: 2022
- Pages: 32
- DOI: https://doi.org/10.5089/9798400224713.001
- Stock No: WPIEA2022259
- ISBN: 9798400224713
- ISSN: 1018-5941
- Main objective: Introduce deep reinforcement learning (DRL) and various DRL algorithms; overview literature on DRL in economics with focus on macromodeling; analyze potentials and limitations of DRL in macroeconomics; identify issues to be addressed for wider DRL adoption in macro modeling.
Theoretical introduction and methods
- Provides a theoretical introduction to deep reinforcement learning and various DRL algorithms.
- Covers DRL algorithmic foundations relevant for economic modeling and decision processes.
- Emphasizes terminology used in the literature: Deep reinforcement learning, DRL, DRL algorithm, Reinforcement learning, RL, RL algorithm, RL algorithm overview, trust region policy optimization.
Literature overview and main applications in macroeconomics
- Surveys recent works demonstrating how DRL can be used to study:
- Optimal policy-making
- Game theory
- Bounded rationality
- Focuses on the main applications of deep reinforcement learning in macromodeling and related econometric and general equilibrium frameworks.
Potentials, limitations, and issues for adoption
- Potentials:
- DRL offers a framework to model complex decision processes and adaptive agents in macroeconomic contexts.
- Capable of addressing problems in optimal policy-making, strategic interactions, and bounded rationality.
- Limitations and issues to be addressed for wider use in macro modeling:
- Methodological and implementation barriers inherent to DRL approaches in macroeconomic models.
- Need for further work to reconcile DRL algorithms with econometric analysis and general equilibrium models.
- Practical challenges in integrating DRL into policy analysis and asset and liability management applications.
Subject areas and keywords
- Subject: Artificial intelligence, Asset and liability management, Debt relief, Econometric analysis, General equilibrium models, Machine learning, Technology
- Keywords: Artificial intelligence, Artificial intelligence, Debt relief, decision process, Deep reinforcement learning, DRL, DRL algorithm, General equilibrium models, Global, Learning algorithms, Machine learning, Macro modeling, Reinforcement learning, RL, RL algorithm, RL algorithm overview, trust region policy optimization
IMF Working Paper No. 2022/259 — "Deep Reinforcement Learning: Emerging Trends in Macroeconomics and Future Prospects", December 16, 2022.
Content in this bundle
- Working Paper