Evolutionary Algorithms
Evolutionary algorithms are a class of optimization algorithms inspired by biological evolution, such as natural selection and genetic inheritance. They maintain a population of candidate solutions, iteratively selecting the 'fittest' ones to 'reproduce' (combine features) and 'mutate' (introduce random changes), gradually evolving towards better solutions for complex problems. These algorithms do not require the objective function to be differentiable, making them versatile.
It's like breeding the best racehorses over generations, allowing slight variations and selecting the fastest offspring to continue the line.
They can find robust solutions to optimization problems that are too complex or ill-defined for traditional mathematical methods.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free