Multi-branch Spatio-Temporal Graph Neural Network for Predicting Ice Layer Thickness from Radar Data

Core Concepts

This research proposes a novel multi-branch spatio-temporal graph neural network that outperforms existing models in accurately and efficiently predicting ice layer thickness from airborne radar data by leveraging GraphSAGE for spatial feature learning and temporal convolution for capturing temporal changes.

Abstract

Liu, Z., & Rahnemooonfar, M. (2024). Multi-branch Spatio-Temporal Graph Neural Network For Efficient Ice Layer Thickness Prediction. Tackling Climate Change with Machine Learning: workshop at NeurIPS 2024. arXiv:2411.04055v1 [cs.LG].

This paper aims to improve the accuracy and efficiency of ice layer thickness prediction using a novel multi-branch spatio-temporal graph neural network. The researchers sought to address the limitations of existing fused spatio-temporal graph neural networks, which suffer from high computational costs and long training times.

Key Insights Distilled From

Multi-branch Spatio-Temporal Graph Neural Network For Efficient Ice Layer Thickness Prediction

by Zesheng Liu,... at arxiv.org 11-07-2024

https://arxiv.org/pdf/2411.04055.pdf