Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 47 for “"Quasi-Newton Methods"”.
-
Quasi-Newton Methods for Nonlinear Programming
… a new class of symmetric updates for use in a quasi-Newton method for nonlinear programming. We show how these updates model the underlying nonlinear equation better than the standard symmetric updates and also how they require less overall work for large problems.
-
SMOOTH QUASI-NEWTON METHODS FOR NONSMOOTH OPTIMIZATION
The success of Newton’s method for smooth optimization, when Hessians are available, motivated the idea of quasi-Newton methods, which approximate Hessians in response to changes in gradients and result in superlinear convergence on smooth functions. Sporadic informal observations over several …
-
SELF-SCALING PARALLEL QUASI-NEWTON METHODS FOR NONLINEAR UNCONSTRAINED OPTIMIZATION
Quasi-Newton (QN) methods are generally considered to be the most efficient algorithms for solving medium-sized problems (with 100 - 500 variables). Currently there are many different updates existed in the literature. But none of them alone can be efficiently applied to solve all types of …
-
A General Convergence Theory for Quasi-Newton Methods for Constrained Optimization
In this thesis we study the local convergence of quasi-Newton methods for nonlinear optimization problems with nonlinear equality constraints. A general theory for analyzing the local convergence of the sequence {x(,k)} generated by the diagonalized quasi-Newton method is developed. Conditions on …
-
On the use of function values to improve Quasi-Newton methods
… constraints or the problem may be unconstrained. Methods of optimization may be classified as derivative (or gradient) methods and non-derivative methods. We assume that the n-variable function to be optimized and its first partial derivatives are all continuous. In this work, we intend to examine …
-
Non-asymptotic study of quasi-Newton methods with application in large-scale machine learning
Quasi-Newton methods are among popular optimization algorithms for solving convex programs since they serve as a middle ground between first-order methods and second-order Newton-type algorithms. They improve the linear convergence rate of first-order methods and achieve a local superlinear rate, …
-
On the Use of Quasi-Newton Methods for the Minimization of Convex Quadratic Splines
In reformulating a strictly convex quadratic program with simple bound constraints as the unconstrained minimization of a strictly convex quadratic spline, established algorithms can be implemented with relaxed differentiability conditions. In this work, the positive definite secant update method …
-
Analytical study of multi-secant quasi-Newton methods for optimization problems : approaching the multi-secant property in symmetric quasi-Newton update formulas
Quasi-Newton (QN) methods are widely used both in root-finding of nonlinear numerical problems and in the search for optimal solutions for applications in industry or research and development. Since the algebraic forms of these problems are often not available, the Jacobian matrix is replaced by an …
-
Quasi-Newton and Multigrid Methods for Semiconductor Device Simulation
… convergence analysis of Gummel's method and quasi-Newton methods is extended to a nonuniform mesh and the Bernoulli function discretization. It is proved that Gummel's method and the quasi-Newton methods for the scaled carrier densities and carrier densities converge locally for sufficiently …
-
Graphical model structure learning using L₁-regularization
… L₁-regularization. We describe limited-memory quasi-Newton methods to solve optimization problems with these types of regularizers, and we examine learning directed acyclic graphical models with L₁-regularization, learning undirected graphical models with group L₁-regularization, and learning …
-
First order sampling methods in gravitational wave parameter estimation
… with ease. Ensemble-based first-order sampling methods offer several notable advantages over traditional approaches. First, they are more robust in exploring energy landscapes and discovering multiple disconnected energy minima. Second, they can be combined with tempering or birth-death …
-
Local and Global Convergence for Convex-Composite Optimization
… data analysis and machine learning. Almost all methods for solving this problem involve direction finding subproblems based on linearizing the smooth function c at some current iterate. When h is the identity function on the real line, these direction finding subproblems correspond to steepest …
-
An Examination of the Strengths and Weaknesses of Newton's Method for Nonlinear Optimization
… linear and nonlinear optimization. While other methods are mentioned, the focus is on analytical methods used to solve nonlinear optimization problems. We briefly look at some of the most effective constrained methods for nonlinear optimization and then show how unconstrained methods often play …
-
Designing optimization algorithms through the lens of online learning
… benign and losses vary gradually, optimistic methods can exploit this predictability to achieve improved regret bounds. Our key observation is that the online losses induced by smooth optimization problems are inherently structured: smoothness naturally induces the type of predictability that …
-
Proximal Methods for Nonconvex Composite Optimization Problems
… function. In this thesis several numerical methods for solving such problems in finite-dimensional spaces are discussed, which are based on proximity operators. After some basic results from convex and nonsmooth analysis are summarized, a first-order method, the proximal gradient method, is …
-
Robust estimation of parameters in nonlinear subsurface flow models using adjoint state methods.
… groundwater flow models by automatic calibration methods is an extremely difficult problem, but one which must be solved in order to produce reliable model predictions. The data upon which the model is calibrated are usually corrupted by measurement and model structure errors which can unduly …
-
Efficient Methods For Large-Scale Empirical Risk Minimization
… begin by focusing on incremental and stochastic methods which split the training samples into smaller sets across time to lower the computation burden of traditional descent algorithms. We develop and analyze convergent stochastic variants of quasi-Newton methods which do not require computation …
-
Numerical investigation of micro-macro coupling in magneto-impedance sensors for weak field measurements
… relied primarily on trial and error experimental methods along with decoupled models that separate the micromagnetic and classical electromagnetic equations describing the MI effect. To offer a basis for more cost-effective designs, this thesis research presentation begins with a general …
Page 1 of 3