The authors lift input features and labels into three-dimensional latent tensors, avoiding the fixed feature dimensions that often bind graph models to a particular dataset schema.
One classifier unifies predictions across nodes, links and whole graphs
The authors map heterogeneous graph features, labels and task types into a shared similarity-based prediction framework.
Big Tech
Ben Finkelshtein · André Linhares · Petar Veličković · Bryan Perozzi · Mikhail Galkin
Google Research · Google DeepMind
Research Digest··2 min read
Finkelshtein et al.
Why this paper
From Google Research and Google DeepMind
In one line
The Universal Classifier unifies node, edge, and graph tasks under a similarity-based model that transfers zero-shot across graphs.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§