Download presentation
Presentation is loading. Please wait.
1
Easy Optimization Problems, Relaxation, Local Processing for a single variable
3
Multiscale solvers Coarsening: create a hierarchy of problems graphs, equations, systems of particles, etc.
4
Original system 1 st coarsening 2 nd coarsening 3 rd coarsening
5
Multiscale solvers Coarsening: create a hierarchy of problems graphs, equations, systems of particles, etc. Solve the coarsest level
6
Coarsest level solution
7
Multiscale solvers Coarsening: create a hierarchy of problems graphs, equations, systems of particles, etc. Solve the coarsest level Uncoarsening: Initialize the solution on a finer level from the coarser level by interpolation Improve the initial solution by local processing
8
Local processing Main assumption The solution of the larger scales has been obtained by the coarser levels At each level apply only local changes Since done iteratively, need not solve to the optimum, just approach it
9
Variable by variable strict unconstrained minimization Discrete (combinatorial) case : Ising model
10
2D Ising spins Minimize Periodic boundary condition Initialize randomly: with probability.5
12
Exc#1: 2D Ising spins exercise Minimize Periodic boundary condition Initialize randomly: with probability.5 1.Go over the grid in lexicographic order, for each spin choose 1 or -1 whichever minimizes the energy (choose with probability ½ when the two possibilities have the same energy) until no changes are observed. 2. Repeat 3 times for each of the 4 possibilities of (h 1,h 2 ). 3. Is the global minimum achievable? 4. What local minima do you observe?
13
Variable by variable strict unconstrained minimization Discrete (combinatorial) case : Ising model Quadratic case : P=2
14
Necessary optimality conditions Let be a local minimum of and assume is continuously differentiable in some domain, then the 1 st order Necessary Condition is If in addition is twice continuously differentiable within, then the 2 nd order Necessary Condition is positive semidefinite
15
Sufficient optimality conditions Let be twice continuously differentiable in domain and let satisfy the conditions, positive definite then is a strict unconstrained local minimum of. If, in addition, is quadratic, the local minimum is also the global unique minimum.
16
Pointwise relaxation for P=2 Minimize Pick a variable, fix all at Minimize Quadratic functional in one variable – easy to solve!
17
Pointwise relaxation for P=2 (cont.) Check the 2 nd derivative: => Unique minimum! Put at the weighted average location of its graph neighbors Go over all variables in lexicographic order Problem: Does not preserve the volume demands! Reinforce volume demands at the end of each sweep
18
Variable by variable strict unconstrained minimization Discrete (combinatorial) case : Ising model Quadratic case : P=2 General functional : P=1, P>2
19
Exc#2: Pointwise relaxation for P=1 Minimize Pick a variable, fix all at Minimize Find the optimal location for
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.