Gradient descent follows the regularization path for general losses.

Ziwei Ji, Miroslav Dudk, Robert E. Schapire, Matus Telgarsky

Browse the full COLT paper archive.