In overparameterized quadratic networks, one-pass SGD escapes generalization plateaus only modestly faster and selects the initialization-closest zero-loss solution due to a conserved quantity in the overlap ODEs.
Neural networks can learn representations with gradient descent
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
cond-mat.dis-nn 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Escape dynamics and implicit bias of one-pass SGD in overparameterized quadratic networks
In overparameterized quadratic networks, one-pass SGD escapes generalization plateaus only modestly faster and selects the initialization-closest zero-loss solution due to a conserved quantity in the overlap ODEs.