8 Appendix 8.1 Proof of proposition 1 L (A) = L0 θ (A) + γL

Neural Information Processing Systems 

Equation (8) uses the linearity of expectation. Includes 100 synthetic 2D graphs, regular tiling of the 2D plane with equilateral triangles, with 100 |V | < 400 [47]. Includes 100 synthetic graphs with 10 |V | 100. Includes 100 synthetic 2D graphs, regular tiling of the 2D plane with equilateral squares, with 100 |V | < 400 [47]. Figure 5: Each dataset is split into train, validation and test sets.