fitness function

In evolutionary computationevolutionary computation, n.: Problem-solving by simulated evolution, much as breeders improve crops: keep many candidate answers, let the better ones produce slightly altered offspring, drop the worse, and repeat for many rounds., the function that assigns each candidate solution a score of its quality, determining its chance of being selected for reproduction; it plays the role of the objective functionobjective function, n.: The score that an optimization process tries to push as low, or as high, as possible. For an AI, it defines what 'doing well' means during training. in optimization or of the rewardreward, n.: The score an AI system receives after an action, telling it how good the outcome was. Reinforcement learning trains the system to collect as much reward as possible over time. in reinforcement learningreinforcement learning, n.: A way of teaching a computer program by trial and error: it tries actions, earns points for good results and penalties for bad ones, and gradually learns which choices pay off over time.. Fitness may be computed analytically, by simulation, by physical experiment or by human judgment (interactive evolution), and is often the most expensive part of the algorithm, which motivates cheap surrogate models. The fitness landscape — fitness viewed as a surface over the space of solutions — frames search difficulty in terms of peaks, plateaus, ruggedness and deception. A poorly specified fitness function gets exploited in unintended ways, a long-documented failure that is an early form of specification gamingspecification gaming, n.: When an AI meets the letter of its goal while defeating its purpose, like a genie who grants exactly the words of a wish instead of what the wisher meant..

The robots learned to flip over and crawl on their elbows, which our fitness function happened to reward.