Physics-Embedded Neural Networks: Graph Neural PDE Solvers with Mixed Boundary Conditions

1 code implementation • 24 May 2022 • Masanobu Horie, Naoto Mitsume

We present our approach termed physics-embedded neural networks that considers boundary conditions and predicts the state after a long time using an implicit method.

Isometric Transformation Invariant and Equivariant Graph Convolutional Networks

1 code implementation • ICLR 2021 • Masanobu Horie, Naoki Morita, Toshiaki Hishinuma, Yu Ihara, Naoto Mitsume

Graphs are one of the most important data structures for representing pairwise relations between objects.

Object Detection Physical Simulations