A custom convergence measure applied to a 96k-parameter decoder-only transformer shows learning-rate boundaries between convergent and divergent training are fractal-like across zoom levels.
Fractal geometry: mathematical foundations and applications
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Mapping the Edge of Chaos: Fractal-Like Boundaries in The Trainability of Decoder-Only Transformer Models
A custom convergence measure applied to a 96k-parameter decoder-only transformer shows learning-rate boundaries between convergent and divergent training are fractal-like across zoom levels.