on optimization problems
TRANSCRIPT
On Optimization Problems
Vahid Moosavi
Supervisor: Ludger HovestadtITA, CAAD
August 2011
Categorization of Optimization Problems based on two main factors
Features, Properties (elements and relations)
and their Values
Desired Output Quality (Desired output Value)
(desired Situation)
Number of …
Uncertainty
Vagueness
Features, Properties (elements and relations)
and their Values
Desired Output Quality (Desired output Value)
(desired Situation)
Unknown,Disagreement
Known,Agreement
Known,Agreement
Unknown,Disagreement
Number of …
Uncertainty
Vagueness
2
Features, Properties (elements and relations)
and their Values
Desired Output Quality (Desired output Value)
(desired Situation)
Unknown,Disagreement
Known,Agreement
Known,Agreement
Unknown,Disagreement
Number of …
Uncertainty
Vagueness
1
3
Infeasible Space
1
Examples:•Designing a precise! ruler for drawing •Gear Box Design•A lot of classic Optimization Problems:• Knapsack Problem• Blending Problem• Classic Shortest path
Problem
Solution Methods:•Classic Optimization Methods (OR)• Optimizing the target
(objective ) functions subject to constraints
Class 1’s Characteristics:•Clear Objective(s) , known Desired output Value, known desired Situation•Clear operational system model
Examples:•Real transportation problems•Communication Systems and (Function Approximation)•Gene Expression problem•Complex machining process modeling
Solution Methods:•Assume Clear system models and then transforming it to class 1 problems•Generalizing Classic Optimization Methods (OR)• Stochastic Optimization • Model (mathematical )free
(Simulation based) •Precise data modeling•Feature selection and remodeling•Problem Restructuring
Semi-Solution Methods:•System Understanding and analysis• System Dynamics
modeling• Cognitive maps • Simulation methods
2Class 2’s Characteristics:
•Clear Objective(s) , known Desired output Value , known desired Situation•Unclear operational system model• Large number of elements and properties• Uncertain elements and relations
Examples:•Messy (wicked) problems•Good health care system•Good City •Good government•Urban Sustainability
Solution Methods:1. Assume Clear Objectives
and then transforming it to class 2 problems
2. Try to model the complex behaviors (e.g. Agent based models in movement of residents)
3. Problem Restructuring 4. Participatory and
collaborative design and planning methods
5. City game and mechanism design
6. Open source technology
Semi-Solution Methods:•System Understanding and analysis• System Dynamics modeling• Cognitive maps • Simulation methods
•Policy analysis
3 Class 3’s Characteristics:•Unclear Objective(s) , Unknown Desired Output Value , Unknown desired Situation•Unclear operational system model• Large number of elements and properties• Uncertain elements and relations
•Stakeholders (humans) are playing in the system•Self-referential systems
Some times having a solution doesn’t make sense, because there
is no clear desired outcomes in reality!!
Thank you
21 3
Centralized model
Decentralized model
Distributed problem solving
Open Source
Centralized model
Decentralized model
Distributed problem solving
Clear desired Solution
Clear desired Solution
(sometimes)
No Clear desired Solution can be
defined
Directing the system instead
of managing