On the exponentially weighted aggregate with the Laplace prior (Q1800807)

This paper studies the exponentially weighted aggregate (EWA) in the context of high-dimensional regression with fixed design and under a sparsity scenario. Oracle inequalities for the EWA with Laplace prior are established. If the temperature parameter is small, the EWA with the Laplace prior satisfies the same type of oracle inequality as the Lasso estimator does, as long as the quality of estimation is measured by the prediction loss. Extensions of the proposed methodology to the problem of prediction with low-rank matrices are obtained.