Teaching Economics to the Machines
Organizers
Speaker
Time
Friday, November 28, 2025 4:00 PM - 5:30 PM
Venue
Shuangqing-B627
Online
Zoom 230 432 7880
(BIMSA)
Abstract
While structural models in economics can offer valuable insights, they often suffer from a poor fit with the data and demonstrate suboptimal forecasting performances. Machine learning models, in contrast, offer rich flexibility but are prone to overfitting and tend to struggle to generalize beyond the confines of training data. We propose a novel framework that incorporates economic restrictions from a structural model into a machine learning model through transfer learning. Specifically, we first construct a neural-network representation of the structural model by training it on the synthetic data generated by the structural model, and then fine-tune the network using real data. In an application to option pricing, the transfer learning model significantly outperforms both the structural model and a conventional data-driven deep neural network. The out-performance is more significant when the sample size of real data is small or under volatile market conditions.
Speaker Intro
Ke Tang is a Professor at the Institute of Economics, School of Social Sciences, and the Dean of Zhishan College at Tsinghua University. His main research interests are Commodity Markets (including Digital Assets), Fintech, and the Digital Economy. He has published numerous high-quality academic papers in the Journal of Finance, Review of Financial Studies, Management Science, PNAS. He serves as the Executive Editor of Quantitative Finance. His research has been recognized by the U.S. Commodity Futures Trading Commission, the United Nations Commodity Report, and various media outlets. He was also selected as a Highly Cited Researcher in China by Elsevier from 2020 to 2023.