The full architecture of CGNN is presented at [CGNN's official site](https://tony-y.github.io/cgnn/architectures/).

Interpolation between [exponential decay](https://paperswithcode.com/method/exponential-decay) and [cosine annealing](https://paperswithcode.com/method/cosine-annealing).

Cosine Power Annealing

sharpDARTS: Faster and More Accurate Differentiable Architecture Search

CGNN

Crystal Graph Neural Networks for Data Mining in Materials Science

[Laplacian eigenvectors](https://paperswithcode.com/paper/laplacian-eigenmaps-and-spectral-techniques) represent a natural generalization of the [Transformer](https://paperswithcode.com/method/transformer) positional encodings (PE) for graphs as the eigenvectors of a discrete line (NLP graph) are the cosine and sinusoidal functions. They help encode distance-aware information (i.e., nearby nodes have similar positional features and farther nodes have dissimilar positional features).

Hence, Laplacian Positional Encoding (PE) is a general method to encode node positions in a graph. For each node, its Laplacian PE is the k smallest non-trivial eigenvectors.

Source	Crystal Graph Neural Networks for Data Mining in Materials Science
Year	2000
Data Source	Papers with Code archive (CC BY-SA)