Deep learning methods for solving dynamic economic models offer computational advantages but produce ‘black box’ solutions. This paper develops Economics Informed Neural Networks (EINNs) through a generalized method of moderation that embeds behavioral bounds directly into neural network architectures.
Novel framework combining macroeconomic and agent-based financial market modeling to study financial stability with household agents.
Efficient method for solving heterogeneous agent models with multiple decisions, extending the Endogenous Grid Method with Gaussian Process Regression for unstructured grids.
Enhances heterogeneous agent models by integrating life cycle considerations and wealth-dependent utility into structural estimation.
Uses a unique stock market model to show how investors with different expectations impact financial stability. Supported by DARPA.