Seminar by Prof. Dacheng Xiu from Econometrics and Statistics at Booth School of Business, University of Chicago

Autoencoders are neural networks widely used in unsupervised learning for dimensionality reduction and feature extraction. This paper provides non-asymptotic guarantees for deep autoencoders within a nonlinear factor model, showing they can effectively extract latent components with errors that diminish with increasing dimensionality and sample size. The extracted factors converge to the true latent factors, up to a functional transformation. We extend these results to supervised autoencoders, supporting their use in factor-augmented prediction and structured matrix completion. Finally, we illustrate the practical value of autoencoders in macroeconomic forecasting, asset return prediction, and noise reduction for causal analysis.

Dacheng Xiu is a professor of Econometrics and Statistics at Booth School of Business, University of Chicago. He serves as a Research Associate at the National Bureau of Economic Research. He currently holds and has previously held several editorial positions, including Co-Editor of Journal of Business & Economic Statistics and Journal of Financial Econometrics, as well as Associate Editor for journals such as Journal of Finance, Review of Financial Studies, Journal of the American Statistical Association, Management Science, and Journal of Econometrics. He has received several recognitions for his research, including Fellow of the Society for Financial Econometrics, Fellow of the Journal of Econometrics, Swiss Finance Institute Outstanding Paper Award, AQR Insight Award, and best paper prizes at various conferences. He has been recognized as one of Poets & Quants' Best 40-under-40 Business School Professors.