I find it alarming how often claims about the performance of graph neural networks tend to crumble upon closer inspection.

The paper below shows how simple greedy algorithms can outperform GNNs by a factor of 10⁴.

@tiago I'm not. Scientists are prone to use their golden hammer on everything. Heck, that's what reviewers and appointment committees want! No one ever asks whether the method made sense.

Of course that means using the golden hammer on screws, electronics, cloth, and not to forget: for surgery.

I'm not invested enough to read in depth, but I do see that the original paper compared against some other algorithms; are those similarly inflated? did the greedy algorithm(s) not appear elsewhere for comparison at some point?

Sign in to participate in the conversation
Mastodon @

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!