Implicit bias in overparameterized models emerges as a geometric correction induced by gradient noise and loss symmetries, enabling inverse design of desired biases like sparsity.
Dynamics in deep classifiers trained with the square loss: Normalization, low rank, neural collapse, and generalization bounds.Research, 6:0024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Understanding and inverse design of implicit bias in stochastic learning: a geometric perspective
Implicit bias in overparameterized models emerges as a geometric correction induced by gradient noise and loss symmetries, enabling inverse design of desired biases like sparsity.