1: \begin{abstract}

2: Fully nonparametric methods for regression from functional data have

3: poor accuracy from a statistical viewpoint, reflecting the fact that

4: their convergence rates are slower than nonparametric rates for the

5: estimation of high-dimensional functions. This difficulty has led to

6: an emphasis on the so-called functional linear model, which is much

7: more flexible than common linear models in finite dimension, but

8: nevertheless imposes structural constraints on the relationship

9: between predictors and responses. Recent advances have extended the

10: linear approach by using it in conjunction with link functions, and

11: by considering multiple indices, but the flexibility of this

12: technique is still limited. For example, the link may be modeled

13: parametrically or on a grid only, or may be constrained by an

14: assumption such as monotonicity; multiple indices have been modeled

15: by making finite-dimensional assumptions. In this paper we introduce

16: a new technique for estimating the link function nonparametrically,

17: and we suggest an approach to multi-index modeling using adaptively

18: defined linear projections of functional data. We show that our

19: methods enable prediction with polynomial convergence rates. The

20: finite sample performance of our methods is studied in simulations,

21: and is illustrated by an application to a functional regression problem.

22: \end{abstract}