On the Benefits of Large Learning Rates for Kernel Methods

Open in new window