Prediction bounds for (higher order) total variation regularized least squares
share
We establish oracle inequalities for the least squares estimator f̂ with penalty on the total variation of f̂ or on its higher order differences. Our main tool is an interpolating vector that leads to lower bounds for compatibility constants. This allows one to show that for any N ∈N the N^ th order differences penalty leads to an estimator f̂ that can adapt to the number of jumps in the (N-1)^ th order differences.
READ FULL TEXT