BIMSA >
Data Analysis and Problem Solving Seminar
Data Analysis and Problem Solving Seminar
Adjoint Gradient Computation for Initial State and Network Parameters in Neural ODE Method
Adjoint Gradient Computation for Initial State and Network Parameters in Neural ODE Method
Organizer
Speaker
Time
Friday, May 15, 2026 3:00 PM - 4:00 PM
Venue
A3-1-301
Online
Zoom 204 323 0165
(BIMSA)
Abstract
The training of Neural ODEs requires the estimations of the initial state as well as the network parameters. This talk introduces the adjoint gradient method for their computations. I will first describe the Neural ODE method for the state evolution in continuous time, and then the adjoint gradient method. A one-dimensional example is presented to illustrate how adjoint variables are propagated backward when losses exist at multiple observation times, and how the initial state and model parameters are optimized accordingly.
Speaker Intro
Zhang Qian is a second-year Ph.D. student in a joint Program between BIMSA and Renmin University of China, majoring in Statistics and Big Data, under the supervision of Professor Zhang Xiaoming. Her current research focuses on the data-driven discovery of differential equations from noisy and irregularly sampled time-series data using machine learning techniques.