Learning rates for the kernel regularized regression with a differentiable strongly convex loss