Abstract

We study the statistical-computational tradeoffs in a high dimensional single index model Y , where f is unknown, X is a Gaussian vector and  is s-sparse with unit norm. When shows that the direction and support of can be recovered using a generalized version of Lasso. In this paper, we investigate the case when this critical assumption fails to hold, where the problem becomes considerably harder. Using the statistical query model to characterize the computational cost of an algorithm, we show that when , no computationally tractable algorithms can achieve the information-theoretic limit of the minimax risk. This implies that one must pay an extra computational cost for the nonlinearity involved in the model.