Differentiable Economics: Strategic Behavior, Mechanisms, and Machine Learning.

Differentiable economics applies neural networks and machine learning to reformulate economic models as differentiable structures, allowing the use of gradient-based optimization. This framework addresses the difficulty of finding Bayes–Nash equilibria in auctions and market games by combining reinf...

Descripción completa

Detalles Bibliográficos
Publicado en:	Communications of the ACM Vol. 68; no. 9; pp. 80 - 89
Autores principales:	Bichler, Martin, Parkes, David C.
Formato:	Artículo
Publicado:	Association for Computing Machinery Sep2025
Materias:	Differentiable functions Economic models Machine learning Artificial neural networks Optimization algorithms Game theory
Acceso en línea:	Ver este registro en EBSCOhost

Descripción
Sumario:	Differentiable economics applies neural networks and machine learning to reformulate economic models as differentiable structures, allowing the use of gradient-based optimization. This framework addresses the difficulty of finding Bayes–Nash equilibria in auctions and market games by combining reinforcement learning with self-play, thus overcoming traditional computational obstacles in game theory. Beyond equilibrium analysis, the approach also enables automated mechanism design, yielding revenue-maximizing auctions and other solutions that are inaccessible through purely analytical methods.