claim
active
claim:spd-avoids-shrinkage-because-importance-norm-penalizes-mask-probability-not-singular-valuesSPD avoids shrinkage because importance norm penalizes mask probability, not singular values
Mechanism.
Source paper
extracted_from(2025) · Lucius Bushnaq · Dan Braun · Lee Sharkey
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- SPD avoids shrinkage.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- SPD's robustness comes from trained importance functions, gradient flow, no simplicity loss, and direct causal importancehypothesis0.795Four hypothesized sources.
- Motivation.
- Central claim.
- Fig. 7d,e.
- Systematic underestimation of feature activations degrades reconstruction and interpretability.
- Discrete case.
- Better bases may give sparsity.
- Systematic underestimation of non-zero feature activations due to L1 sparsity penalty.