Evolutionary computation is a family of problem-solving techniques in artificial intelligence inspired by the process of natural evolution. Instead of following a strict, step-by-step algorithm to find a solution, evolutionary computation methods work by simulating populations of possible solutions that evolve over time. This approach is especially useful for tackling complex problems where traditional algorithms struggle, such as optimization tasks with large, hard-to-navigate search spaces or problems where it’s difficult to define an exact solution path.
The foundational idea of evolutionary computation is to use mechanisms similar to those found in biological evolution: selection, mutation, recombination (or crossover), and survival of the fittest. Algorithms in this family maintain a population of candidate solutions. Each generation, solutions are evaluated using a fitness function that measures how well they solve the problem. The best-performing individuals are selected to produce the next generation, typically by combining features from two or more parent solutions (crossover) and introducing small random variations (mutation). Over many generations, the population “evolves” toward better solutions.
The most common types of evolutionary computation methods include genetic algorithms, genetic programming, evolutionary strategies, and evolutionary programming. While these approaches differ in some technical details, they share the core philosophy of using population-based, stochastic (randomized) search guided by a fitness function. For example, in a genetic algorithm, candidate solutions are often represented as strings of numbers or symbols (like a chromosome), and the process mimics natural genetics. In evolutionary strategies, solutions may be represented as vectors of real numbers, with a focus on self-adaptation of mutation rates.
Evolutionary computation is valued for its flexibility and robustness. Unlike many traditional algorithms, it does not require the problem to be differentiable, continuous, or even fully understood. This makes it applicable to real-world scenarios such as scheduling, engineering design, game playing, and machine learning parameter tuning. However, evolutionary algorithms can be computationally intensive, as they often need to evaluate many candidate solutions over many generations. Parallel processing and hybrid methods are sometimes used to speed up the search.
A key advantage of evolutionary computation is its ability to escape local optima. Traditional optimization algorithms can become trapped in suboptimal solutions, but the random mutation and recombination steps in evolutionary computation help maintain diversity in the population. This diversity increases the chances of discovering better solutions that might be overlooked by more deterministic approaches.
Overall, evolutionary computation is a core technique within computational intelligence and remains an active area of research and practical application. It continues to inspire new algorithms and hybrid methods for solving challenging problems across science, engineering, and beyond.
