Theory-guided Learning of Deep Model for Partial Differential Equations

Description

Research project at Graduate School of Information and Physical Science at Osaka University. This is a collaborated work with Department of Architecture for Intelligence, The Institute of Scientific and Industrial Research(ISIR), Osaka University. This research is the approach to supervising neural networks by specifying constraints that should hold over the output space by combining complex physical model, rather than direct examples of input-output pairs.

Conferences and Workshops

• Ken-ichi Fukui, Junya Tanaka, Tomohiko Tomita, and Masayuki Numao. "Physics-guided Neural Network with Model Discrepancy Based on Upper Troposphere Wind Prediction", *Proc. IEEE 18th International Conference on Machine Learning and Applications (ICMLA 2019), Dec. 2019.*
• Junya Tanaka, Ken-ichi Fukui, Tomohiko Tomita, and Masayuki Numao. "Learning Model Discrepancy in Physics-guided Neural Networks", The 33rd Annual Conference of the Japanese Society for Artificial Intelligence, June. 2019.
• 田中 潤也, 冨田 智彦, 沼尾 正行, 福井 健一. “物理法則に基づくニューラルネットワーク構築の検討 ー対流圏上層の風予測を例にー”, 情報処理学会第117回数理モデル化と問題解決研究発表会, Mar. 2018.(研究奨励賞)