REGULARIZED NEWTON METHOD WITH GLOBAL O (1/k2) CONVERGENCE

被引:14
作者
Mishchenko, Konstantin [1 ]
机构
[1] CNRS, Ecole Normale Super, Inria, F-75005 Paris, France
关键词
second-order optimization; Newton's method; Levenberg-Marquardt algorithm; global convergence; LEVENBERG-MARQUARDT METHOD; CUBIC REGULARIZATION; LINE SEARCH; COMPLEXITY; MINIMIZATION;
D O I
10.1137/22M1488752
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
We present a Newton-type method that converges fast from any initialization and for arbitrary convex objectives with Lipschitz Hessians. We achieve this by merging the ideas of cubic regularization with a certain adaptive Levenberg-Marquardt penalty. In particular, we show that the iterates given by x(k+1) = x(k) (del(2)f(x(k)) + root H parallel to del f(x(k))parallel to I)(-1) del f(x(k)), where H > 0 is a constant, converge globally with a O (1/k(2)) rate. Our method is the first variant of Newton's method that has both cheap iterations and provably fast global convergence. Moreover, we prove that locally our method converges superlinearly when the objective is strongly convex. To boost the method's performance, we present a line search procedure that does not need prior knowledge of H and is provably efficient.
引用
收藏
页码:1440 / 1462
页数:23
相关论文
共 72 条
[1]   Oracle complexity of second-order methods for smooth convex optimization [J].
Arjevani, Yossi ;
Shamir, Ohad ;
Shiff, Ron .
MATHEMATICAL PROGRAMMING, 2019, 178 (1-2) :327-360
[3]   A CONTINUOUS DYNAMICAL NEWTON-LIKE APPROACH TO SOLVING MONOTONE INCLUSIONS [J].
Attouch, H. ;
Svaiter, B. F. .
SIAM JOURNAL ON CONTROL AND OPTIMIZATION, 2011, 49 (02) :574-598
[4]  
Attouch H, 2016, J CONVEX ANAL, V23, P139
[5]  
Baes M., 2009, ESTIMATE SEQUENCE ME
[6]   GFN2-xTB-An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions [J].
Bannwarth, Christoph ;
Ehlert, Sebastian ;
Grimme, Stefan .
JOURNAL OF CHEMICAL THEORY AND COMPUTATION, 2019, 15 (03) :1652-1671
[7]  
Bechtle S., 2020, Conference on Robot Learning, P162
[8]   Convergence and Complexity Analysis of a Levenberg-Marquardt Algorithm for Inverse Problems [J].
Bergou, El Houcine ;
Diouane, Youssef ;
Kungurtsev, Vyacheslav .
JOURNAL OF OPTIMIZATION THEORY AND APPLICATIONS, 2020, 185 (03) :927-944
[9]   THE USE OF QUADRATIC REGULARIZATION WITH A CUBIC DESCENT CONDITION FOR UNCONSTRAINED OPTIMIZATION [J].
Birgin, E. G. ;
Martinez, J. M. .
SIAM JOURNAL ON OPTIMIZATION, 2017, 27 (02) :1049-1074
[10]   Curiosities and counterexamples in smooth convex optimization [J].
Bolte, Jerome ;
Pauwels, Edouard .
MATHEMATICAL PROGRAMMING, 2022, 195 (1-2) :553-603