Optimal-transport alignment of adjacent layers gives a cheap and slightly better initialization for progressive depth up-scaling than copying, averaging, or a learned predictor.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Progressive Depth Up-scaling via Optimal Transport
Optimal-transport alignment of adjacent layers gives a cheap and slightly better initialization for progressive depth up-scaling than copying, averaging, or a learned predictor.