A population is a collection of candidate solutions in one generation of an evolutionary algorithm. Think of it as several attempts at a puzzle being tested at once. Each candidate, or individual, is a chromosome made of genes. Keeping varied candidates lets the search explore different parts of a large solution space instead of relying on one attempt. section 5.22 section 4.2 section 4.4
A population of knapsacks
Suppose a knapsack has a 9 kg weight limit and eight available items. Represent each candidate with eight binary digits: 1 means include an item; 0 means leave it out. One candidate includes items 1, 3, and 7, weighing 9 kg and worth $14. Other strings represent other packing choices. Together, those strings form the population. section 4.2 section 4.4
A population changes as the algorithm runs:
- Create an initial set of varied, random candidates. Check the problem’s constraints: an invalid candidate can be repaired, discarded, or given a poor fitness score. section 4.3 section 4.6
- Score each individual with a fitness function. For a knapsack, higher total value is better only if the weight limit is respected; an overweight candidate can receive a score of zero. section 4.7
- Select parents, produce offspring, then choose who enters the next generation. Population size is usually fixed, so adding offspring means removing some candidates. section 4.3 section 4.17
How many individuals are replaced is a design choice. A steady-state model replaces a portion each generation; a generational model replaces everyone. Keeping only the highest-scoring individuals, called elitism, can reduce diversity if those survivors are too similar. A varied population matters especially when a search might otherwise get stuck at a local best, though evaluating many candidates takes computation. section 4.9 section 4.17 section 4.2
def generate_initial_population(population_size):
population = []
for individual in range(0, population_size):
individual = ''.join([random.choice('01')
➥for n in range(26)]) #A
population.append([individual, 0, 0]) #B
return population
This Python example builds an initial population by generating random 26-digit binary strings and storing each as an individual. It shows population creation, but does not itself check whether a candidate satisfies knapsack constraints. section 4.6
