Global Optimization General issues in global optimization Classification of algorithms The DIRECT algorithm – Relationship to EGO – Lipschitzian optimization.

Slides:



Advertisements
Similar presentations
Rachel T. Johnson Douglas C. Montgomery Bradley Jones
Advertisements

Cognitive Radio Communications and Networks: Principles and Practice By A. M. Wyglinski, M. Nekovee, Y. T. Hou (Elsevier, December 2009) 1 Chapter 12 Cross-Layer.
Computational Intelligence Winter Term 2009/10 Prof. Dr. Günter Rudolph Lehrstuhl für Algorithm Engineering (LS 11) Fakultät für Informatik TU Dortmund.
Application a hybrid controller to a mobile robot J.-S Chiou, K. -Y. Wang,Simulation Modelling Pratice and Theory Vol. 16 pp (2008) Professor:
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION GROUP
Space-Filling DOEs These designs use values of variables inside range instead of at boundaries Latin hypercubes, one popular space- filling DOE uses as.
Sampling plans for linear regression
 Will help you gain knowledge in: ◦ Improving performance characteristics ◦ Reducing costs ◦ Understand regression analysis ◦ Understand relationships.
Unsupervised Learning Clustering K-Means. Recall: Key Components of Intelligent Agents Representation Language: Graph, Bayes Nets, Linear functions Inference.
Graphical optimization Some problems are cheap to simulate or test. Even if they are not, we may fit a surrogate that is cheap to evaluate. Relying on.
Lecture #3; Based on slides by Yinyu Ye
Towards a Theoretic Understanding of DCEE Scott Alfeld, Matthew E
Classification / Regression Support Vector Machines
Coverage by Directional Sensors Jing Ai and Alhussein A. Abouzeid Dept. of Electrical, Computer and Systems Engineering Rensselaer Polytechnic Institute.
Monte Carlo Methods and Statistical Physics
Engineering optimization dilemma Optimization algorithms developed by mathematicians are normally based on linear and quadratic approximations Usually.
Global Optimization General issues in global optimization Classification of algorithms The DIRECT algorithm – Divided rectangles – Exploration and Exploitation.
1 Learning with continuous experts using Drifting Games work with Robert E. Schapire Princeton University work with Robert E. Schapire Princeton University.
Optimal Design Laboratory | University of Michigan, Ann Arbor 2011 Design Preference Elicitation Using Efficient Global Optimization Yi Ren Panos Y. Papalambros.
Including Uncertainty Models for Surrogate-based Global Optimization
Geo479/579: Geostatistics Ch14. Search Strategies.
Tea Break!.
Speeding up multi-task learning Phong T Pham. Multi-task learning  Combine data from various data sources  Potentially exploit the inter-relation between.
Petroleum Reservoir Management Based on Approximate Dynamic Programming Zheng Wen, Benjamin Van Roy, Louis Durlofsky and Khalid Aziz Smart Field Consortium,
Copyright 2004, K. Leoviriyakit and A. Jameson Challenges and Complexity of Aerodynamic Wing Design Kasidit Leoviriyakit and Antony Jameson Stanford University.
1 Multi-point Wing Planform Optimization via Control Theory Kasidit Leoviriyakit and Antony Jameson Department of Aeronautics and Astronautics Stanford.
MAE 552 – Heuristic Optimization Lecture 4 January 30, 2002.
Statistics: The Science of Learning from Data Data Collection Data Analysis Interpretation Prediction  Take Action W.E. Deming “The value of statistics.
MAE 552 – Heuristic Optimization Lecture 1 January 23, 2002.
Applications in GIS (Kriging Interpolation)
Lecture 9 – Nonlinear Programming Models
Metaheuristics The idea: search the solution space directly. No math models, only a set of algorithmic steps, iterative method. Find a feasible solution.
Space-Filling DOEs Design of experiments (DOE) for noisy data tend to place points on the boundary of the domain. When the error in the surrogate is due.
Sachin Kumar Porandla Advisor Dr. Wenzhong Gao Optimization of Hybrid Powertrain using DIRECT Algorithm POWERTRAIN DESIGN GROUP MEETING #5.
1 CE 530 Molecular Simulation Lecture 7 David A. Kofke Department of Chemical Engineering SUNY Buffalo
Trajectory Correction and Tuning James Jones Anthony Scarfe.
Group Recommendations with Rank Aggregation and Collaborative Filtering Linas Baltrunas, Tadas Makcinskas, Francesco Ricci Free University of Bozen-Bolzano.
EQT373 STATISTIC FOR ENGINEERS Design of Experiment (DOE) Noorulnajwa Diyana Yaacob School of Bioprocess Engineering Universiti Malaysia Perlis 30 April.
1 IE 607 Heuristic Optimization Particle Swarm Optimization.
Monte Carlo Methods1 T Special Course In Information Science II Tomas Ukkonen
Ensemble Based Systems in Decision Making Advisor: Hsin-His Chen Reporter: Chi-Hsin Yu Date: IEEE CIRCUITS AND SYSTEMS MAGAZINE 2006, Q3 Robi.
1 汽车优化设计 王琥 湖南大学 机械与运载工程学院 汽车车身先进设计制造国家重点实验室.
Map of the Great Divide Basin, Wyoming, created using a neural network and used to find likely fossil beds See:
11C. Linear Programming. What is a linear programming problem? 1.A set of variables (in Further Maths there will only ever be two variables) called decision.
Linear Programming Advanced Math Topics Mrs. Mongold.
Progress Report Qiang Chen, Derek Dalle, Chad Griep, Jingwei Hu, Jahmario Williams, Zhenqiu Xie Multiobjective Modeling and Optimization in Design.
1 Efficient experimentation for nanostructure synthesis using Sequential Minimum Energy Designs (SMED) V. Roshan Joseph +, Tirthankar Dasgupta* and C.
Machine Learning Queens College Lecture 7: Clustering.
Monte-Carlo based Expertise A powerful Tool for System Evaluation & Optimization  Introduction  Features  System Performance.
INCLUDING UNCERTAINTY MODELS FOR SURROGATE BASED GLOBAL DESIGN OPTIMIZATION The EGO algorithm STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION GROUP Thanks.
Evolutionary Computing Chapter 12. / 26 Chapter 12: Multiobjective Evolutionary Algorithms Multiobjective optimisation problems (MOP) -Pareto optimality.
Bayesian Optimization. Problem Formulation Goal  Discover the X that maximizes Y  Global optimization Active experimentation  We can choose which values.
Kriging - Introduction Method invented in the 1950s by South African geologist Daniel Krige (1919-) for predicting distribution of minerals. Became very.
Graphical optimization Some problems are cheap to simulate or test. Even if they are not, we may fit a surrogate that is cheap to evaluate. Relying on.
COLLECTING AND PROCESSING OF INFORMATION PRESENTATION © 2011 International Technology and Engineering Educators Association, STEM  Center for Teaching.
Geometric Approach Geometric Interpretation:
Multi-Scale Search for Black-Box Optimization: Theory & Algorithms
March 9th, 2015 Maxime Lapalme Nazim Ould-Brahim
Homework Bell Work: Agenda Friday August 11, 2017 Objectives:
Map of the Great Divide Basin, Wyoming, created using a neural network and used to find likely fossil beds See:
Collecting and processing of information Presentation 4.5.1
Matching of Propulsion Systems for an Aircraft
AE 440 Performance Discipline Lecture 9
Chapter 14.
Graphical optimization
Meta-Heuristic Algorithms 16B1NCI637
3-3 Optimization with Linear Programming
Collecting and processing of information Presentation 4.5.1
Collecting and processing of information Presentation 4.5.1
Structural Optimization Design ( Structural Analysis & Optimization )
Presentation transcript:

Global Optimization General issues in global optimization Classification of algorithms The DIRECT algorithm – Relationship to EGO – Lipschitzian optimization – Exploration and Exploitation – Application to High Speed Civil Transport

Global optimization issues Optimization problem is NP-hard No-free-lunch theorem (Wolpert and Macready) – No single algorithm can do well on all problems – If an algorithm is improved for one problem, it will suffer for others. Great opportunity for engineers to use problem knowledge to tailor algorithms. Big headache for journals because they get many worthless new algorithms.

Global Optimization Global optimization algorithms by Thomas Weise

Global trend If there is global trend, the only issue is not to be trapped in a local minimum If there are many important basins there is no choice but to explore thoroughly.

global optimization algorithms

EGO and DIRECT EGO is an algorithm that relies on a surrogate to connect all the data in design space. DIRECT is an algorithm that does not try to exploit any correlation between the function value at one point and its value at nearby points. EGO will do well for functions with global trend and not too many deep local optima. DIRECT will do well for cheap functions with many local optima. We combine them by using DIRECT to calculate new sampling points for EGO.

DIRECT Algorithm Jones, D.R., Perttunen, C.D. and Stuckman, B.E. (1993), Lipschitzan optimization without the Lipschitz constant, Journal of Optimization Theory and Application 79, 157–181. S. E. COX, R. T. HAFTKA, C. A. BAKER, B. GROSSMAN, W. H. MASON and L. T. WATSON A Comparison of Global Optimization Methods for the Design of a High-speed Civil Transport, Journal of Global Optimization 21: 415–433, Matlab:

Lipschitzian Optimization Optimizer divides space into boxes and samples the vertices of each One box is further divided based on a predicted maximum rate of change of the function, K

DIRECT The function value at the middle of each box and its largest dimension are used to determine potentially optimal boxes Each potentially optimal box is divided Lipschitzian optimization for all possible Lipschitz constants

DIRECT Box Division.

Exploration vs. Exploitation DIRECT uses convex hull of box sizes to balance exploitation vs. exploration. With enough function evaluations every region in design space will be explored. This is clearly not feasible for high dimensional spaces. Coxs paper compares DIRECT to repeated local optimization with random starts.

Optimization of a High Speed Civil Transport 26 design variables defining the shape of the wing and fuselage and tail and the flight trajectory Long range (NY Tokyo type flight) with 250 passengers for minimum takeoff gross weight Constraints on takeoff and landing maneuvers, engine out conditions. CFD analysis plus structural optimization needed for each design.

Results