Summary

• GAs randomized parallel hill-climbing search for hypotheses that optimize a predefined fitness function
• based on analogy to biological evolution - diverse population of competing hypotheses, at each iteration most fit members of the population are selected, combined by crossover and subjected to random mutation
• GAs show how learning can be seen as a special case of optimization - learning task is finding optimal hypothesis - this suggests other optimization techniques - like simulated annealing - can be applied to machine learning