Regularization and the small-ball method. I: Sparse recovery (Q1750281): Difference between revisions

Let \((\Omega, \mu)\) be a probability space and set \(X\) to be distributed according to \(\mu\). \(F\) is a class of real-valued functions defined on \(\Omega\), \(Y\) is the unknown random variable that one would like to approximate using function in \(F\) and \(\lambda\) is the regularization parameter. The authors discuss the best approximation to \(y\) and find the function \(f^*\) that minimizes in \(F\) the squared loss functional \(f\to E(f(x)-y)^2\).