HORST uses non-commutative operator composition and a hyperbolic mirror map to combine stability from adaptive optimizers with L1 sparsity bias, outperforming AdamW across sparsity levels on vision and language tasks.
The Combinatorial Brain Surgeon: Pruning Weights That Cancel One Another in Neural Networks , publisher =
1 Pith paper cite this work, alongside 4 external citations. Polarity classification is still indexing.
1
Pith paper citing it
4
external citations · external index
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
HORST: Composing Optimizer Geometries for Sparse Transformer Training
HORST uses non-commutative operator composition and a hyperbolic mirror map to combine stability from adaptive optimizers with L1 sparsity bias, outperforming AdamW across sparsity levels on vision and language tasks.