What graph neural networks cannot learn: depth vs width

Authors: Andreas Loukas

ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Section 5 presents some empirical evidence of the theory. This section aims to empirically test the connection between the capacity dw of a GNNmp, the number of nodes n of its input, and its ability to solve a given task.
Researcher Affiliation Academia Andreas Loukas Ecole Polytechnique Fédérale Lausanne andreas.loukas@epfl.ch
Pseudocode Yes Computational model 1 Message passing graph neural network (GNNmp)
Open Source Code No The paper does not contain any explicit statement about open-sourcing the code or providing a link to a code repository.
Open Datasets No I generated five distributions over graphs with n (8, 16, 24, 32, 44) nodes and an average diameter of (4, 6, 8, 9, 11), respectively (this was achieved by setting p (6, 8, 10, 12, 14), see Figure 3a). For each such distribution, I generated a training and test set consisting respectively of 1000 and 200 examples.
Dataset Splits No For each such distribution, I generated a training and test set consisting respectively of 1000 and 200 examples. Both sets were exactly balanced, i.e., any example graph from the training and test set had exactly 50% chance of containing a 4-cycle.
Hardware Specification No The paper does not specify any hardware details such as GPU models, CPU types, or other computing resources used for experiments.
Software Dependencies No The GNNmp chosen was that proposed by Xu et al. (2018), with the addition of residual connections this network outperformed all others that I experimented with. The paper mentions using Adam optimizer and learning rate decay, but does not specify version numbers for any software or libraries.
Experiment Setup Yes To this end, I performed grid search over the hyperparameters w (2, 10, 20) and d (5, 10, 20, 15). To reduce the dependence on the initial conditions and training length, for each hyperparameter combination, I trained 4 networks independently (using Adam and learning rate decay) for 4000 epochs.