Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably

Headnotes

Many equations and formulas look intimidating. However, when you hunt them down, they are definitely not! Just papertigers!
Now let's hunt the papertiger.

Graph neural networks

 (1)    \begin{equation*} x^{t} = \text{Update}( x_{\text{current}}^{t-1},  \square_{\text{all ajacent nodes}} \text{Message}(x_{\text{current}}^{t-1}, x_{\text{ajacent}}^{t-1)}, \text{edge feature})) \end{equation*}
 (2)    \begin{equation*} H^{(t+1)} = F(H^t, X) \end{equation*}

One single big graph

Many nonisomorphic graphs

Many isomorphic graphs

Footnotes

There are many excellent tutorials out there. Some tutorials are too intuitive and it's helpful, but you cannot get it straight on the math details. Some focused on dymestifying math. Some focused on code. I found the best tutorials that give you the conceptual ideas and are possible for implementation without being blind to the math details. Drop a comment if I failed. It would be really appreciable.

If you want to cite this article, please cite this article as:

Lachlan Chen, "Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably," in EarnFromScratch, September 11, 2020, https://www.earnfs.com/en/html/2617.htm.

or

@misc{lachlanchen2020tutorial,
title=Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably,
author={Chen, Lachlan},
year=September 11, 2020
}


EarnFromScratch (March 2, 2021) Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably. Retrieved from https://www.earnfs.com/en/html/2617.htm.
"Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably." EarnFromScratch - March 2, 2021, https://www.earnfs.com/en/html/2617.htm
EarnFromScratch September 11, 2020 Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably., viewed March 2, 2021,<https://www.earnfs.com/en/html/2617.htm>
EarnFromScratch - Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably. [Internet]. [Accessed March 2, 2021]. Available from: https://www.earnfs.com/en/html/2617.htm
"Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably." EarnFromScratch - Accessed March 2, 2021. https://www.earnfs.com/en/html/2617.htm
"Hunt the papertiger of graph neural networks (GNNs), intuitively, mathematically, implementably." EarnFromScratch [Online]. Available: https://www.earnfs.com/en/html/2617.htm. [Accessed: March 2, 2021]

Leave a Reply