09/30/2026
π’ New Publication from Prof. Hiroyuki Fujiwara and collaborators!
Weβre pleased to highlight a new publication featuring Prof. Hiroyuki Fujiwara and researchers from Gifu University, J.A. Woollam Co., J.A. Woollam Japan, ROHM, and AIST:
π¬ βAdaptive mixture-of-experts denoising for deep learning of optical ellipsometry spectraβ
Published in Machine Learning: Science and Technology, this work presents a deep-learning approach that combines a Mixture-of-Experts (MoE) architecture with a U-Net model to adaptively reduce noise in spectroscopic ellipsometry data.
The researchers demonstrate the approach using experimental data from 27 semiconductor thin-film materials measured with two different J.A. Woollam ellipsometry systems. The iterative MoE approach reduced global DLE prediction error by up to 65.0% for 1-second measurements and improved bandgap estimation accuracy by 44.3%.
π₯ Authors:
Yuki Yamamoto, Ren Iwayama, Yuki Ikarashi, James N. Hilfiker, Jyunya Takashima, Takuji Maekawa, Hitoshi Tampo, and Hiroyuki Fujiwara.
Congratulations to Prof. Fujiwara and the entire research team on this work advancing AI-assisted spectroscopic ellipsometry and high-throughput materials characterization!
π https://iopscience.iop.org/article/10.1088/2632-2153/ae9691