Title: Spectral estimators for multi-index models Authors:  Yihan Zhang - University of Bristol (United Kingdom) [presenting]
Abstract: Multi-index models provide a popular framework to investigate the learnability of functions with low-dimensional structure. Due to their connections with neural networks, they have been an object of recent intensive study. The focus is on recovering the subspace spanned by the signals via spectral estimators -- a family of methods that are routinely used in practice, often as a warm-start for iterative algorithms. The main technical contribution is a precise asymptotic characterization of the performance of spectral methods, when sample size and input dimension grow proportionally and the dimension $p$ of the space to recover is fixed. Specifically, we locate the top-$p$ eigenvalues of the spectral matrix and establish the overlaps between the corresponding eigenvectors (which give the spectral estimators) and a basis of the signal subspace. Our analysis unveils a phase transition phenomenon in which, as the sample complexity grows, eigenvalues escape from the bulk of the spectrum and, when that happens, eigenvectors recover directions of the desired subspace. The precise characterization we put forward enables the optimization of the data preprocessing, thus allowing us to identify the spectral estimator that requires the minimal sample size for weak recovery.