1: \begin{abstract}

2: \noindent The multi-index model with sparse dimension reduction matrix

3: is a popular approach to circumvent the curse of dimensionality in

4: a high-dimensional regression setting. Building on the single-index

5: analysis by Alquier, P. \& Biau, G. (Journal of Machine Learning Research

6: 14 (2013) 243--280), we develop a PAC-Bayesian estimation method

7: for a possibly misspecified multi-index model with unknown active

8: dimension and an orthogonal dimension reduction matrix. Our main result

9: is a non-asymptotic oracle inequality, which shows that the estimation

10: method adapts to the active dimension of the model, the sparsity of

11: the dimension reduction matrix and the regularity of the link function.

12: Under a Sobolev regularity assumption on the link function the estimator

13: achieves the minimax rate of convergence (up to a logarithmic factor)

14: and no additional price is paid for the unknown active dimension.

15: \end{abstract}

16: