rademacher complexity
Rademacher complexity is a measure of the capacity of a class of functions to fit random noise. It is used in statistical learning theory to derive bounds on the generalization error of learning algorithms, helping to understand their potential performance on unseen data.
- Block Coordinate Descent for Neural Networks Provably Finds Global Minima
- Dynamical Decoupling of Generalization and Overfitting in Large Two-Layer Networks
- Dynamical Decoupling of Generalization and Overfitting in Large Two-Layer Networks
- Generalization Bound of Gradient Flow through Training Trajectory and Data-dependent Kernel
- Multiclass Loss Geometry Matters for Generalization of Gradient Descent in Separable Classification
- Optimal Rates for Generalization of Gradient Descent for Deep ReLU Classification
- Reproducing Kernel Banach Space Models for Neural Networks with Application to Rademacher Complexity Analysis