Deep Reinforcement Learning in a Monetary Model
Prof. Mingli Chen
Associate Professor
Department of Economics
University of Warwick
We propose deep reinforcement learning (DRL) as a general approach to modeling bounded rationality in dynamic stochastic general equilibrium (DSGE) frameworks. Agents are represented by deep artificial neural networks and learn to maximize their intertemporal objective functions by interacting with an a priori unknown environment. Applying this approach to a model from the adaptive learning literature, DRL agents can learn all equilibria, regardless of local stability properties. However, learning can be slow and may become unstable without the use of early stopping criteria. These findings have implications for both the interpretation of DRL agents and the use of DSGE models more broadly.
