A selective sparsity regularizer, SPR, nudges link prediction metrics up on standard benchmarks, but the evidence is single-run and the theoretical justification is mathematically flawed.
No Reg") quickly achieved near-perfect scores on training data (MRR and Hits@1 close to 1.0). However, their performance on validation set data (
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Rethinking Regularization Methods for Knowledge Graph Completion
A selective sparsity regularizer, SPR, nudges link prediction metrics up on standard benchmarks, but the evidence is single-run and the theoretical justification is mathematically flawed.