Understanding Genetic Algorithms

3 min readMay 10, 2021

Understanding aspects of evolutionary programming with genetic algorithms

Introduction

Genetic algorithms are general-purpose search and optimization algorithms that use principles inspired by natural population genetics to evolve solutions to problems
GA is based on evolutionary programming and uses stochastic gradient optimization
GAs operate on a population of individuals representing potential solutions to a given problem.
GAs seek to produce better (fitter) individuals (solutions) by combining the better of the existing ones (through breeding -crossover and mutation).

Initially we define our problem and our objective function.
We define our initial population size.
We perform genetic operations like selection, crossover and mutation on the selected population to get fitter genes for the next generation.

Selection-focuses attention on high fitness individuals, thus exploiting on the available fitness information. High-performing structures might be chosen several times for replication, while poor-performing structures might not be chosen at all.
Crossover-Combines the features of two parent structures (new combinations of genes are generated from previous ones) to form two similar offspring.
Mutation-Alters one or more components of a selected structure randomly (a direct analogy from nature and provides the way to introduce new information into the population).
This process goes on till we reach a global optimal value with the best possible fitness scores for the genes.

First, we initialize a random sample space for our initial population. It should represent a random sample of the search space. Each member of the population is evaluated and assigned a measure of its fitness as a solution.
Members are then selected based on the fitness function defined by us. Structures in the current population are selected for replication based on their relative fitness. Genetic operators (crossover, mutation) are performed to generate the new population.
The resulting offspring are then evaluated and inserted back into the population, replacing older members.
Specific decisions about how many members are replaced during each iteration, and how members are selected for replacement, define a range of alternative implementations
Generational GA-GAs that replace the entire population
Steady-state GA-GAs that replace only a small fraction of chromosomes. Typically new chromosomes replace the worst chromosomes

I have used the ‘geneticalgorithm’ library to write the objective function.
You can install the library using the pip package-‘pip install geneticalgorithm’
My code can be found here-https://github.com/Doraiswamy/Adaptive-Test/blob/master/SystemCode/backend/faculty/GA.py

Combinatorial optimization
Resource allocation problems
Classical problem of travelling salesperson / bin-packing / job shop scheduling etc.
Financial forecasting
Design optimization-network routing, Satellite orbit selection etc.
Machine learning loss function optimization