Lecture derives R-GCN's 4.75M-parameter bill and a 14.7× basis-decomposition fix
JeremyCMorgan · x · 2026-11-06
A GNN course lecture on heterogeneous graphs: it hand-derives a 4.75M-parameter relational GCN to show how relation types explode model complexity, then covers basis and block-diagonal decompositions (basis trick cuts parameters 14.7×), attention-based models (HAN metapaths, HGT typed attention), and an honest encoder–decoder KG-completion showdown (R-GCN + DistMult vs shallow embeddings, same decoder and eval). Useful math before committing to a GNN architecture.
More from Research
- UC Berkeley's Ctrl-F adds force commands so humanoid robots can move 20 kg loads — ChongZzZhang · 2026-11-06
- Archon Robotics debuts Anchor 1, a humanoid whole-body controller cutting tracking error 30-80% — ChongZzZhang · 2026-11-06
- Steering one attention head flips Qwen3-30B from 'I am an LLM' to 'I am a real person' — Sauers_ · 2026-11-06
- Stanford 2.5-hour lecture explains end-to-end how ChatGPT and Claude are built — MaxLi_Columbia · 2026-11-06
- Formal proofs of tens of millions of lines of code are now feasible, says Anthropic engineer — morqon · 2026-11-06
- Hugging Face releases FinePhrase slides: 486B-token synthetic pretraining dataset with 90 rewrite styles — NielsRogge · 2026-11-06