Conjugate Gradient Optimization

Slides:



Advertisements
Similar presentations
Instabilities of SVD Small eigenvalues -> m+ sensitive to small amounts of noise Small eigenvalues maybe indistinguishable from 0 Possible to remove small.
Advertisements

Optimization.
1 OR II GSLM Outline  some terminology  differences between LP and NLP  basic questions in NLP  gradient and Hessian  quadratic form  contour,
Introducción a la Optimización de procesos químicos. Curso 2005/2006 BASIC CONCEPTS IN OPTIMIZATION: PART II: Continuous & Unconstrained Important concepts.
Least Squares example There are 3 mountains u,y,z that from one site have been measured as 2474 ft., 3882 ft., and 4834 ft.. But from u, y looks 1422 ft.
Function Optimization Newton’s Method. Conjugate Gradients
Tutorial 12 Unconstrained optimization Conjugate gradients.
AppxA_01fig_PChem.jpg Complex Numbers i. AppxA_02fig_PChem.jpg Complex Conjugate.
Curve Fitting and Interpolation: Lecture (II)
ECIV 301 Programming & Graphics Numerical Methods for Engineers REVIEW II.
Monica Garika Chandana Guduru. METHODS TO SOLVE LINEAR SYSTEMS Direct methods Gaussian elimination method LU method for factorization Simplex method of.
CSE 245: Computer Aided Circuit Simulation and Verification
Many quadratic equations can not be solved by factoring. Other techniques are required to solve them. 7.1 – Completing the Square x 2 = 20 5x =
7.1 – Completing the Square
9.4 – Solving Quadratic Equations By Completing The Square
Section 8.3 – Systems of Linear Equations - Determinants Using Determinants to Solve Systems of Equations A determinant is a value that is obtained from.
# 1 # 2 # 3 Multiply # 4 Multiply: Multiply # 5 # 6 Solve.

9 1 Performance Optimization. 9 2 Basic Optimization Algorithm p k - Search Direction  k - Learning Rate or.
Example: Introduction to Krylov Subspace Methods DEF: Krylov sequence
UNCONSTRAINED MULTIVARIABLE
MAT 2401 Linear Algebra 2.3 The Inverse of a Matrix
AppxA_01fig_PChem.jpg Complex Numbers i. AppxA_02fig_PChem.jpg Complex Conjugate * - z* =(a, -b)
Solving Quadratic Equations by Factoring. Solution by factoring Example 1 Find the roots of each quadratic by factoring. factoring a) x² − 3x + 2 b) x².
1 Incorporating Iterative Refinement with Sparse Cholesky April 2007 Doron Pearl.
CSE 245: Computer Aided Circuit Simulation and Verification Matrix Computations: Iterative Methods I Chung-Kuan Cheng.
Data Modeling Patrice Koehl Department of Biological Sciences National University of Singapore
Warm-up 1.What are the different ways to solve quadratic equation? Solve the following problem by completing the square.
Section Day 2 Solving Trig Equations Using Inverses.
Solve by factoring. x² = - 4 – 5x 2,. Solve by factoring. n² = -30 – 11n -4 and -1.
Signal & Weight Vector Spaces
DIFFERENTIAL EQUATIONS Note: Differential equations are equations containing a derivative. They can be solved by integration to obtain a general solution.
Improving Performance of The Interior Point Method by Preconditioning Project Proposal by: Ken Ryals For: AMSC Fall 2007-Spring 2008.
Title: ch. 6 Algebra: quadratic equations. Learning Objective: WALT solve quadratic equations by using the quadratic formula. p.143.
Factoring to Solve Quadratic Equations – Solving Quadratic Equations by Factoring A quadratic equation is written in the Standard Form, where a,
The Landscape of Sparse Ax=b Solvers Direct A = LU Iterative y’ = Ay Non- symmetric Symmetric positive definite More RobustLess Storage More Robust More.
Zero-Offset Data d = L o ò r ) ( g = d dr r ) ( g = d
Solving Quadratic Equations by the Complete the Square Method
Computational Optimization
Quasi-Newton Methods Problem: SD, CG too slow to converge if NxN H matrix is ill-conditioned. SD: dx = - g (slow but no inverse to store or compute) QN:
LECTURE 3 OF SOLUTIONS OF NON -LINEAR EQUATIONS.
Ch. 5 – Applications of Derivatives
Vector Calculus and Quadratic function
CSE 245: Computer Aided Circuit Simulation and Verification
Iterative Non-Linear Optimization Methods
Steepest Descent Optimization
CS5321 Numerical Optimization
5.6 – The Quadratic Formula And Ch 5 Review
Class Notes 11.2 The Quadratic Formula.
Conjugate Gradient Problem: SD too slow to converge if NxN H matrix is ill-conditioned. SD: dx = - g (slow but no inverse to store or compute) CG: dx =
Collaborative Filtering Matrix Factorization Approach
CSE 245: Computer Aided Circuit Simulation and Verification
CS5321 Numerical Optimization
9.4 Solving Quadratic Equations
CS5321 Numerical Optimization
Conjugate Gradient Method
CS5321 Numerical Optimization
Solving Quadratics Using Square Roots
~ Least Squares example
Quadratic Equations.
4.5: Linear Approximations, Differentials and Newton’s Method
~ Least Squares example
Bellwork: 2/19/18 1) Factor and solve: 3x2_7x-20=0
Administrivia: November 9, 2009
Performance Optimization
Outline Preface Fundamentals of Optimization
Steepest Descent Optimization
Computer Animation Algorithms and Techniques
Conjugate Direction Methods
First-Order Methods.
Presentation transcript:

Conjugate Gradient Optimization

Outline CG Method Non-linear CG Solving Linear System of Equations Preconditioned CG and Regularization

Quasi-Newton Condition: g’ – g = Hdx’ Recall: d f(x)/dx ~ [df(x+dx)/dx-df(x)/dx]/dx 2 Step 1: Quadratic f(x+dx’) ~ f(x) + dx’ g + 1/2dx’ H dx’ Step 2: Gradient f(x+dx’) ~ g + H dx’ D Step 3: Define f(x+dx’) = g’  g’ – g = Hdx’ D Similar to FD approx. to 2nd dervative x* dx _ g’ g dx’ g’ dx Kiss point x

Conjugate Gradient dx’ = bdx– g’ s.t. (dx’, Hdx)=0 dx _ g D dx Quasi-Newton Condition: g’ – g = Hdx’ (1) Plane spanned by dx and dx’ dx’ x* dx _ g g dx’ dx’ dx’ Kiss point dx Bullseye ( f(x*),dx)=0 D Bullseye ( f(x*),dx’)=0 g’ at bullseye has no components in dx & dx’ plane If dx’ points at bullseye, then dot product of dx with eq. 1 gives -dxg = 0 = dxHdx’ (3) Conjugacy Cond. -> search dx’ (conjugate to previous dx) hits bullseye dx’ = bdx– g’ s.t. (dx’, Hdx)=0 Above is conjugate to dx if b found

{ Conjugate Gradient Conjugacy Cond : dx’Hdx = 0 (3) dx New Search Direc: New Search Direc: dx’ = bdx – g’ (4) { dx’ is conjugate to dx for b s.t. (dx’,Hdx)=0 ( bdx – g’,Hdx)=0 (5) b = g’Hdx/dxHdx (6) Solve for Conjugacy Step b Step length a : a = (g,dx)/dxHdx (7) Hit Bullseye No H!

Conjugate Gradient dx’ For k =2:Niter x* dx No H! end Fletcher-Reeves

Conjugate Gradient For k =2:Niter end dx dx’ Starting point x0 and take -g as initial direction dx’ For k =2:Niter x* dx Polack-Ribiere end

Conjugate Gradient Properties For i = 1:nit %find b p= dx +bg %find a x* g’ dx’ dx’ = dx + ap dx’’ dx= dx’ Kiss point dx x=x+ dx’ end 1. (p ,Hp )=0 for any i=j i j 2. Converges in N steps for NxN H with SPD property 3. Converges quickly if eigenvalues are clustered (i.e., round contours)

Solving Rectangular Linear Systems by Regularized SD with Scaling Given: H rectangular matrix s.t. Hx=g ill-conditioned Let CH H x = Cg s.t. C approximates inverse H H T T x = x – [CH( H x - g) + l Gx (k+1) (k) T (k) Soln: a CG will converge in 2 steps! Newton in one step! 1 1 4 5 2 x1 x2 = Classroom Exercise: Derive formula for b 2. Write CG code and solve above equations

Solving Square Linear Systems by Regularized CG HL Solving Lx = -g  e = xTg + 0.5 xT L x

Solving Rectangular Linear Systems by Regularized CG Solving LTLx = -L Tg  e = xTg + 0.5 xT LTL x Write Two Subroutines: 1). [d]=forward(m) % predict data d=Lm from model 2). [g] = adjoint(r) % predict model g=L’r from data residual r=(Lm-d)