1: \begin{abstract}

2: This article provides convergence analysis of online stochastic gradient

3: descent algorithms for functional linear models.

4: Adopting the characterizations of the slope function regularity, the kernel

5: space capacity, and the capacity of the sampling process covariance operator,

6: significant improvement on the convergence rates is achieved.

7: Both prediction problems and estimation problems are studied,

8: where we show that capacity assumption can alleviate the saturation of the

9: convergence rate as the regularity of the target function increases.

10: We show that with properly selected kernel, capacity assumptions

11: can fully compensate for the regularity assumptions for prediction problems

12: (but not for estimation problems).

13: This demonstrates the significant difference between the prediction problems and the

14: estimation problems in functional data analysis.

15: \end{abstract}

16: