Pure Tensorflow implementation of the SGR layer as proposed in "Symbolic Graph Reasoning Meets Convolutions". This work makes several assumptions as to how the authors thought about certain implementation details. Therefore, in no way is this implementation guaranteed to produce correct or similar results as in the paper.

We additionally add the fasttext embeddings and adjacency matrix used in our experiments,

Liang, Xiaodan, et al. "Symbolic graph reasoning meets convolutions." Advances in Neural Information Processing Systems. 2018.