The performance of many algorithms in the fields of hard combinatorial problem solving, machine learning or AI in general depends on tuned hyperparameter configurations. Automated methods have been proposed to alleviate users from the tedious and error-prone task of manually searching for performance-optimized configurations across a set of problem instances. However there is still a lot of untapped potential through adjusting an algorithm's hyperparameters online since different hyperparameters are potentially optimal at different stages of the algorithm. We formulate the problem of adjusting an algorithm's hyperparameters for a given instance on the fly as a contextual MDP, making reinforcement learning (RL) the prime candidate to solve the resulting algorithm control problem in a data-driven way. Furthermore, inspired by applications of algorithm configuration, we introduce new white-box benchmarks suitable to study algorithm control. We show that on short sequences, algorithm configuration is a valid choice, but that with increasing sequence length a black-box view on the problem quickly becomes infeasible and RL performs better.

From phone apps and GPS maps to music recommendations and artificial intelligence, our lives are increasingly molded by algorithms. Sets of instructions for completing tasks or solving problems, algorithms are the governing principles of our age--the underlying equations that help us make decisions, and, in some cases, make decisions for us. Are these life recipes a force for good or ill? Pew surveyed 1,302 experts of various stripes--futurists, academics, coders, IT guys and girls--and opinions were split. Asked if the "net overall effect" for individuals and society would be positive or negative in the next decade, 38% said the positives would outpace the negatives, while 37% saw it the other way.