B.M.

S College of Engineering

Department of Machine Learning

Nature Inspired Computing
UNIT 03
Genetic Algorithm

Prof. VARSHA R
Asst. Prof, Dept. of MEL, BMSCE
Introduction

Presentation by Prof. Varsha R

How Genetic Algorithm works?

Presentation by Prof. Varsha R

Step 1 Initial Population

Presentation by Prof. Varsha R

Step 2 Fitness Function

Presentation by Prof. Varsha R

Step 3 Selection

Presentation by Prof. Varsha R

Step 3 Selection

Presentation by Prof. Varsha R

Step 4 Crossover

Presentation by Prof. Varsha R

Step 5 Offspring

Presentation by Prof. Varsha R

Step 6 Mutation

Presentation by Prof. Varsha R

Step 6 Mutation

Presentation by Prof. Varsha R

Step 7 Termination

Presentation by Prof. Varsha R

Working of Genetic Algorithm

Presentation by Prof. Varsha R

Pseudo Code for Genetic Algorithm
1) Randomly initialize populations p
2) Determine fitness of population
3) Until convergence repeat:
a) Select parents from population
b) Crossover and generate new population
c) Perform mutation on new population
d) Calculate fitness for new population

Presentation by Prof. Varsha R

Encoding Techniques in Genetic Algorithm

Presentation by Prof. Varsha R

Binary Encoding

Presentation by Prof. Varsha R

Binary Encoding

Presentation by Prof. Varsha R

Binary Encoding Example 1

Presentation by Prof. Varsha R

Binary Encoding Example 2

Presentation by Prof. Varsha R

Value Encoding

Presentation by Prof. Varsha R

Value Encoding Example 1

Presentation by Prof. Varsha R

Permutation (or order) Encoding

Presentation by Prof. Varsha R

Permutation (or order) Encoding Example 1

Presentation by Prof. Varsha R

Tree Encoding

Presentation by Prof. Varsha R

Parent Selection Operators

Presentation by Prof. Varsha R

Fitness Proportionate Selection

Presentation by Prof. Varsha R

Roulette Wheel Selection

Presentation by Prof. Varsha R

Roulette Wheel Selection

Presentation by Prof. Varsha R

Stochastic Universal Sampling (SUS)

Presentation by Prof. Varsha R

Ranking Selection

Presentation by Prof. Varsha R

Ranking Selection

Presentation by Prof. Varsha R

Tournament Selection

Presentation by Prof. Varsha R

Truncation Selection

Presentation by Prof. Varsha R

Crossover Operators

Presentation by Prof. Varsha R

Binary Coded Crossover Operators

Presentation by Prof. Varsha R

Single Point Crossover

Presentation by Prof. Varsha R

Single Point Crossover

Presentation by Prof. Varsha R

Two Point Crossover

Presentation by Prof. Varsha R

Two Point Crossover

Presentation by Prof. Varsha R

Two Point Crossover

Presentation by Prof. Varsha R

Multi Point Crossover

Presentation by Prof. Varsha R

Multi Point Crossover

Presentation by Prof. Varsha R

Multi Point Crossover

Presentation by Prof. Varsha R

Uniform Crossover

Presentation by Prof. Varsha R

Uniform Crossover

Presentation by Prof. Varsha R

Uniform Crossover

Presentation by Prof. Varsha R

Half- Uniform Crossover

Presentation by Prof. Varsha R

Half- Uniform Crossover

Presentation by Prof. Varsha R

Uniform Crossover with Crossover Mask
(CM)

Presentation by Prof. Varsha R

Uniform Crossover with Crossover Mask
(CM)

Presentation by Prof. Varsha R

Uniform Crossover with Crossover Mask
(CM)

Presentation by Prof. Varsha R

Shuffle Crossover

Presentation by Prof. Varsha R

Shuffle Crossover

Presentation by Prof. Varsha R

Shuffle Crossover

Presentation by Prof. Varsha R

Three Parent Crossover

Presentation by Prof. Varsha R

Three Parent Crossover

Presentation by Prof. Varsha R

Three Parent Crossover

Presentation by Prof. Varsha R

Real Coded Crossover Operators

Presentation by Prof. Varsha R

Single Arithmetic Crossover

Presentation by Prof. Varsha R

Single Arithmetic Crossover

Presentation by Prof. Varsha R

Single Arithmetic Crossover

Presentation by Prof. Varsha R

Single Arithmetic Crossover

Presentation by Prof. Varsha R

Linear Crossover

Presentation by Prof. Varsha R

Linear Crossover

Presentation by Prof. Varsha R

Linear Crossover

Presentation by Prof. Varsha R

Linear Crossover

Presentation by Prof. Varsha R

Linear Crossover

Presentation by Prof. Varsha R

Order Coded Crossover Operator

Presentation by Prof. Varsha R

Partially Mapped Crossover

Presentation by Prof. Varsha R

Partially Mapped Crossover

Presentation by Prof. Varsha R

Partially Mapped Crossover

Presentation by Prof. Varsha R

Partially Mapped Crossover

Presentation by Prof. Varsha R

Partially Mapped Crossover

Presentation by Prof. Varsha R

Cycle Crossover

Presentation by Prof. Varsha R

Cycle Crossover

Presentation by Prof. Varsha R

Mutation Operators

Presentation by Prof. Varsha R

Mutation Operators

Presentation by Prof. Varsha R

Bit Flip Mutation

Presentation by Prof. Varsha R

Random Resetting

Presentation by Prof. Varsha R

Swap Mutation

Presentation by Prof. Varsha R

Scramble Mutation

Presentation by Prof. Varsha R

Inversion Mutation

Presentation by Prof. Varsha R

Stopping Condition for Genetic Algorithm
Flow

Presentation by Prof. Varsha R

Stopping Condition for Genetic Algorithm
Flow

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 1

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 2

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 2

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 2

Presentation by Prof. Varsha R

Problem Solving Using GA- Example 2
• Repeat the steps until we get the maximum value for f(x) function.

Presentation by Prof. Varsha R

Classification of GA
1. Messy Genetic Algorithm
2. Adaptive Genetic Algorithm
3. Hybrid Genetic Algorithm
4. Parallel Genetic Algorithm

Presentation by Prof. Varsha R

Messy Genetic Algorithms (mGA)
• Concept: Messy Genetic Algorithms are designed to work efficiently with highly complex
optimization problems by allowing individuals to represent partial solutions. They address issues
related to premature convergence and inefficiency in representing solutions in standard GAs.
• Key Characteristics:
1. Variable-length chromosomes: Unlike standard GAs where chromosome lengths are fixed,
mGAs allow variable lengths to represent solutions.
2. Explicit gene representation: Genes in mGAs include both their position (locus) and value,
enabling a more flexible representation.
3. Building blocks emphasis: mGAs focus on evolving and combining "building blocks" or
smaller problem subcomponents to form complete solutions.

Presentation by Prof. Varsha R

Messy Genetic Algorithms (mGA)
• Phases:
1. Initialization Phase: Starts with a diverse population of chromosomes, including short
(partial) solutions.
2. Juxta positional Phase: Combines short solutions into longer, more comprehensive ones.
3. Clean-up Phase: Refines the solutions by removing redundant or unfit components.
• Applications:
1. Problems with highly interdependent variables.
2. Complex combinatorial optimization tasks, such as scheduling and network design.

Presentation by Prof. Varsha R

Adaptive Genetic Algorithms (AGA)
• Concept: Adaptive Genetic Algorithms dynamically adjust their parameters (e.g., mutation and
crossover rates) during the evolution process to enhance convergence speed and maintain diversity.
• Key Characteristics:
• Dynamic parameter tuning: Adapt mutation and crossover rates based on the performance of the
current population.
• Focus on diversity: Prevents premature convergence by adapting to the search space’s
characteristics.
• Feedback-based adaptation: Uses feedback from fitness values to guide the adjustments.

Presentation by Prof. Varsha R

Adaptive Genetic Algorithms (AGA)
• Advantages:
• Reduces the need for manual parameter tuning.
• Balances exploration (diversity) and exploitation (convergence) effectively.
• Applications:
• Problems where the search space characteristics change during optimization.
• Dynamic optimization problems, such as real-time system control.

Presentation by Prof. Varsha R

Hybrid Genetic Algorithms (hGA)
• Concept: Hybrid Genetic Algorithms combine GAs with other optimization techniques (e.g., local
search, simulated annealing) to leverage the strengths of each method.
• Key Characteristics:
• Enhanced exploration and exploitation: GAs handle global exploration, while other methods
refine local solutions.
• Domain-specific customization: Integrates problem-specific heuristics or algorithms for better
performance.
• Modular design: Combines multiple optimization paradigms flexibly.

Presentation by Prof. Varsha R

Hybrid Genetic Algorithms (hGA)
• Types of Hybrids:
1. GA + Local Search (Memetic Algorithms): Incorporates a local search phase after the
GA operators to fine-tune solutions.
2. GA + Simulated Annealing: Combines the probabilistic nature of simulated annealing
with the population-based approach of GAs.
3. GA + Gradient Descent: Uses gradient-based methods for refining solutions generated
by GAs.
• Applications:
• Large-scale optimization problems, such as vehicle routing and resource allocation.
• Scenarios requiring both high-quality solutions and computational efficiency.

Presentation by Prof. Varsha R

Parallel Genetic Algorithms (pGA)
• Concept: Parallel Genetic Algorithms distribute the computational workload across multiple
processors to accelerate the optimization process and explore the search space more effectively.
• Key Characteristics:
• Distributed population: Divides the population into smaller subpopulations (islands) that evolve
independently.
• Migration mechanism: Periodically exchanges individuals between subpopulations to maintain
diversity.

Presentation by Prof. Varsha R

Parallel Genetic Algorithms (pGA)
• Parallelism levels:
• Coarse-grained parallelism: Subpopulations evolve in parallel on different processors.
• Fine-grained parallelism: Each individual or group of genes is processed in parallel.
• Master-slave parallelism: A central processor (master) manages fitness evaluation
distributed to worker nodes (slaves).
• Advantages:
• Significantly faster convergence, especially for computationally intensive problems.
• Increased diversity due to isolated evolution in subpopulations.
• Scalability with the number of processors.
• Applications:
• High-dimensional optimization problems, such as protein folding and aerodynamic design.
• Scenarios requiring rapid solution times, like real-time simulations.

Presentation by Prof. Varsha R