ARCq: a new adaptive regularization by cubics. (4th March 2018)

• Record Type: Journal Article
• Authors: Dussault, Jean-Pierre
• Is Part Of: Optimization methods and software. Volume 33:Number 2(2018)
• Page Start: 322
• Page End: 335
• Publication Date: 2018-03-04
• Subjects: Nonlinear optimization -- unconstrained optimization -- Trust region algorithms -- Adaptive cubic regularization methods; Mathematical optimization -- Periodicals; Algorithms -- Periodicals; 519.7
• DOI: 10.1080/10556788.2017.1322080
• Languages: English
• ISSNs: 1055-6788

Abstract

Abstract : We present a new adaptive regularization by cubics (ARC) algorithm, which we name ARC and which is very close in spirit to trust region methods, closer than the original ARC is. We prove global convergence to second-order critical points. We also obtain as a corollary the convergence of the original ARC method. We prove the optimal complexity property for the ARC and identify the key elements which allow it. We end by proposing an efficient implementation using a Cholesky like factorization. Limited preliminary experimentation suggests that ARC may be more robust than its trust region counterpart and that our implementation is efficient.