## Abstract

ℓ_{1}-regularized linear inverse problems frequently arise in signal processing, image analysis, and statistics. The correct choice of the regularization parameter t ∈ R ≥0 is a delicate issue. Instead of solving the variational problem for a fixed parameter, the idea of the homotopy method is to compute a complete solution path u(t) as a function of t. In their celebrated paper A new approach to variable selection in least squares problems [IMA J. Numer. Anal. 20 (2000), no. 3, 389-403], Osborne, Presnell, and Turlach showed that the computational cost of this approach is often comparable to the cost of solving the corresponding least squares problem. Their analysis relies on the one-at-a-time condition, which requires that different indices enter or leave the support of the solution at distinct regularization parameters. In this paper, we introduce a generalized homotopy algorithm based on a nonnegative least squares problem, which does not require such a condition, and prove its termination after finitely many steps. At every point of the path, we give a full characterization of all possible directions. To illustrate our results, we discuss examples in which the standard homotopy method either fails or becomes infeasible. To the best of our knowledge, our algorithm is the first to provably compute a full piecewise linear and continuous solution path for an arbitrary combination of a measurement matrix and a data vector.

Original language | English |
---|---|

Pages (from-to) | 2343-2364 |

Number of pages | 22 |

Journal | Mathematics of Computation |

Volume | 87 |

Issue number | 313 |

DOIs | |

State | Published - 2018 |

## Keywords

- Compressed sensing
- Convex optimization
- Homotopy
- Lasso
- Nonnegative least squares
- ℓ-norm
- ℓ-regularization

## Fingerprint

Dive into the research topics of 'The homotopy method revisited: Computing solution paths of ℓ_{1}-regularized problems'. Together they form a unique fingerprint.