DNN classifiers achieve excess risk O(n^{-α}) with α arbitrarily large under hard margin (q=∞) and distribution-adapted smoothness of the regression function, with matching minimax lower bounds for q≥2.
Understanding square loss in training over- parametrized neural network classifiers.arXiv preprint arXiv:2112.03657
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Super-fast Rates of Convergence for Neural Network Classifiers under the Hard Margin Condition
DNN classifiers achieve excess risk O(n^{-α}) with α arbitrarily large under hard margin (q=∞) and distribution-adapted smoothness of the regression function, with matching minimax lower bounds for q≥2.