0% found this document useful (0 votes)
0 views12 pages

Multi Objective Evolutionary Algorithms 2

The document discusses Multi-Objective Evolutionary Algorithms (MOEAs) which address real-world problems involving multiple conflicting goals, such as balancing performance and cost in manufacturing. MOEAs generate a set of trade-off solutions known as the Pareto Optimal Set, providing decision-makers with flexibility in choosing solutions based on their priorities. The document outlines the workings, advantages, limitations, and applications of MOEAs across various industries, emphasizing their role in optimizing complex problems without relying on traditional single-objective algorithms.

Uploaded by

Jyoti Yadav
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
0 views12 pages

Multi Objective Evolutionary Algorithms 2

The document discusses Multi-Objective Evolutionary Algorithms (MOEAs) which address real-world problems involving multiple conflicting goals, such as balancing performance and cost in manufacturing. MOEAs generate a set of trade-off solutions known as the Pareto Optimal Set, providing decision-makers with flexibility in choosing solutions based on their priorities. The document outlines the workings, advantages, limitations, and applications of MOEAs across various industries, emphasizing their role in optimizing complex problems without relying on traditional single-objective algorithms.

Uploaded by

Jyoti Yadav
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Multi-Objective Evolutio ary

Algorit
Balancing Conflicting Goals Using Nature-Inspired Optimization
T e C alle ge of Multiple Goal
Real-world problems rarely have just one objective. Most
involve multiple competing goals that conflict with each
other.

Consider a car manufacturer: they want vehicles that are


faster (high performance) while also being fuel-efficient
(low cost). Improving one often compromises the other.

These are called Multi-Objective Optimization Problems


(MOOPs), and traditional single-objective algorithms fall
short.

MOEAs don't find a single "best" solution4they discover a set of trade-off solutions called the Pareto Optimal Set, giving
decision-makers flexibility to choose based on priorities.
Real-World Applicatio

E gi eeri g De ig Eco o ic Bu i e Mac i e Lear i g E ergy Sy te


Designing airplane wings for Building portfolios with Training models with high Managing power grids for
maximum lift and minimum maximum profit and accuracy but low reliability and efficiency
drag minimum risk computational cost

Healt care
Drug design balancing
effectiveness with minimal
side effects
W y Traditio al Algorit
Fall S ort
Si gle-Objective Li itatio
Traditional algorithms optimize only one goal, failing when
objectives conflict

Populatio -Ba ed Power


MOEAs explore many solutions simultaneously, uncovering diverse
trade-offs

Gradie t-Free
Works on complex, nonlinear problems without requiring
mathematical derivatives

Deci io Flexibility
Provides multiple optimal solutions for decision-makers to
choose from

In manufacturing, MOEAs simultaneously balance production speed,


energy consumption, and machine wear4optimizing the entire
system, not just one metric.
How MOEA Work: Evolutio i Actio
MOEAs follow natural evolution principles4survival of the fittest4to find optimal solutions.

I itializatio
Generate a random population of candidate solutions

Evaluatio
Score each solution across multiple objectives

Selectio
Choose superior solutions based on non-dominance

Cro over
Combine parent solutions to create offspring

Mutatio
Randomly modify solutions to maintain diversity

Replace e t
Keep the best and most diverse solutions for next generation

Co verge ce
Repeat until stopping criteria met4producing a Pareto optimal set
U der ta di g Pareto Opti ality
T e Core Co cept
A solution is Pareto Optimal if no other solution
can improve one objective without worsening
another.

Exa ple Co pari o


Design A: 200 km/h, ¹5 lakh

Design B: 210 km/h, ¹8 lakh

Neither dominates4both are Pareto


optimal

The Pareto Front is the collection of all such


optimal solutions, creating a curve of best trade-
offs.
Key C aracteri tic of MOEA

Populatio -Ba ed Searc No -Do i ated Sorti g Diver ity Mai te a ce


Evaluates many candidate Classifies solutions by Ensures solutions spread across
solutions simultaneously, dominance level4a core the Pareto front using
exploring the solution space mechanism in algorithms like techniques like crowding
comprehensively NSGA-II distance

Eliti Strategy U iver al Adaptability


Preserves the best solutions Works on any problem where
across generations, preventing objectives can be evaluated
loss of quality numerically4no assumptions
required
Leadi g MOEA Algorit

NSGA-II SPEA2
Non-Dominated Sorting Genetic Algorithm II Strength Pareto Evolutionary Algorithm 2

Fast ranking with crowding distance for diversity4the Fitness assignment based on dominance strength and
most widely used MOEA density estimation

MOEA/D PAES
Multi-objective EA based on Decomposition Pareto Archived Evolution Strategy

Breaks complex problems into simpler subproblems for Simple approach using an archive of non-dominated
efficient solving solutions
Adva tage a d Li itatio
' Adva tage ¦ Li itatio

Si ulta eou Opti izatio Co putatio al Co t


Handles multiple conflicting objectives at once Requires many evaluations4can be expensive

Multiple Solutio Para eter Se itivity


Finds diverse trade-offs in a single run Performance depends on tuning population size
and operators

No Mat e atical Co trai t


Deci io Burde
Works without gradients or equations
Provides many solutions4users must choose based
on preferences
Proble Flexibility
Handles discrete, continuous, or hybrid problems
T e Power of Bala ced
Solutio
"There is no single perfect solution4only the best balance among
competing goals."

MOEAs provide a flexible, powerful approach to multi-objective


optimization problems. Inspired by Darwin's theory of evolution, they
evolve diverse populations toward optimal trade-offs.

By producing a Pareto Front4a set of equally good, balanced


solutions4MOEAs empower decision-makers across industries: from
automotive and robotics to finance and healthcare.

In a world of competing priorities, MOEAs find the best path forward.


Vi ual Guide to MOEA

1. MOEA Proce Flow: T e Evolutio ary Cycle


01 02

I itialize Populatio Evaluate Objective


Create a diverse set of initial solutions. Assess each solution against all defined objectives.

03 04

No -Do i ated Sorti g Selectio


Rank solutions based on Pareto dominance to identify Choose the best solutions to become parents for the next
optimal trade-offs. generation.

05 06

Cro over Mutatio New Ge eratio


Generate new solutions by combining and modifying parent Form a new population with improved solutions.
solutions.

07 08

Co verge ce C eck Pareto Fro t


Determine if the population has converged to a stable set Identify the set of non-dominated solutions representing
of solutions. optimal compromises.

2. Pareto Fro t Vi ualizatio

The Pareto Front illustrates the set of optimal trade-offs where no single objective can be improved without sacrificing
another. Solutions on this front are "non-dominated," meaning they are superior to "dominated" solutions, which can be
improved in at least one objective without worsening any other.

3. Key MOEA Applicatio Exa ple

Auto otive De ig Fi a cial Portfolio


Optimizing vehicle weight, fuel efficiency, and safety Balancing risk and return to maximize investment
simultaneously. performance.

Mac i e Lear i g E gi eeri g Opti izatio


Tuning model parameters for accuracy while minimizing Designing structures for strength, cost-effectiveness, and
computational cost. material usage.
T a k You
BY

Harshit Sharma ( 2023UEA4129 )


Aditya Singh ( 2023UEA4206 )
Jyoti Yadav ( 2023UEA4194 )
Harshit ( 2023UEA6601 )

You might also like