Stephen Chen, Princeton University

Abstract

Observations sampled at different frequencies provide multi-level details. How to effectively use them remains a challenging task. We can aggregate all the data to the lowest-frequency level, or we can fill all the missing data to the highest-frequency level. Is there a better way to utilize the information without eliminating the details or introducing noise? We here provide a recurrent neural network model for mixed frequency data, providing convergence proof, error estimation, and some numerical results.

SPECIAL NOTE: This seminar is presented online only.

You can join via ZOOM or by clicking this link https://princeton.zoom.us/j/843514245

For further information see: https://sites.math.rutgers.edu/~zeilberg/expmath/