We go from Message Passing GNN (MPGNN) to TOPOLOGICAL Message Passing on CW Networks: Lifting a Graph to a higher topological space allows for high-dimensional interactions (greater than 2) given our higher-dim topological spaces. Computational Graph Neural Networks increase its complexities and n-body interactions (eg chemistry, pharma, molecular design). We go from Message Passing GNN (MPGNN) to TOPOLOGICAL Message Passing on CW Networks.
First part of this video is here:
https://www.youtube.com/watch?v=Xiy_bD1CHro
All credits go to:
Michael Bronstein, "Beyond Message Passing, a Physics-Inspired Paradigm for Graph Neural Networks", The Gradient, 2022.
P. Veličković, Message passing all the way up (2022) arXiv:2202.11097.
J. Zhu et al., Beyond homophily in graph neural networks: Current limitations and effective designs (2020), NeurIPS.
P. Veličković et al., Graph Attention Networks (2018) ICLR.
Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges by Michael M. Bronstein et al. https://arxiv.org/abs/2104.13478
Weisfeiler and Lehman Go Topological: Message Passing Simplicial Networks
Cristian Bodnar, Fabrizio Frasca, Yu Guang Wang, Nina Otter, Guido Montúfar, Pietro Liò, Michael Bronstein
https://arxiv.org/abs/2103.03212
Just discovered a great explanatory video on topic of CWL by https://www.youtube.com/channel/UCVt9rN2LzKzNwJG-VXvZGkAhttps://www.youtube.com/watch?v=wACDSoDNTfE
#ai
#topologicalspace
#deeplearning
#topology
2 Comments