College of Sciences
In this work, we propose the Variational Autoencoder Inverse Mapper (VAIM) to solve inverse problems, where there is a demand to accurately restore hidden parameters from indirect observations. VAIM is an autoencoder-based neural network architecture. The encoder and decoder networks approximate the forward and backward mapping, respectively, and a variational latent layer is incorporated into VAIM to learn the posterior parameter distributions with respect to the given observables. VAIM shows promising results on several artificial inverse problems. VAIM further demonstrates preliminary effectiveness in constructing the inverse function mapping quantum correlation functions to observables in a quantum chromodynamics analysis of nucleon structure and hadronization.
Download Full Text (6.0 MB)
Almaeen, Manal; Alanazi, Yasir; Kuchera, Michelle; Sato, Nobuo; Melnitchouk, Wally; and Li, Yaohang, "VAIM for Solving Inverse Problems" (2021). College of Sciences Posters. 6.