Solution of linear ill-posed problems by model selection and aggregation (Q1639200)

The authors propose two alternative approaches for overcomplete dictionaries based estimation in linear ill-posed problems. The first estimator is obtained by minimizing penalized empirical risk. The second one is based on a Q-aggregation type procedure that is specifically designed for solutions of linear ill-posed problems. Oracle inequalities are established for both estimators that hold with high probabilities and in expectation. Moreover, for the Q-aggregation estimator, the inequalities are sharp. A performance of the proposed procedures is assessed by a simulation study. All proofs are given. The obtained results are very useful for many applications.